diff --git a/.github/workflows/test.yaml b/.github/workflows/test.yaml new file mode 100644 index 0000000..4fab835 --- /dev/null +++ b/.github/workflows/test.yaml @@ -0,0 +1,15 @@ +name: Check + +on: + pull_request: + branches: [master] + +jobs: + check: + runs-on: ubuntu-22.04 + name: Pre-Commit + steps: + - name: Checkout + uses: actions/checkout@v3 + - name: Run Pre-Commit + uses: pre-commit/action@v3.0.0 diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml new file mode 100644 index 0000000..6938691 --- /dev/null +++ b/.pre-commit-config.yaml @@ -0,0 +1,18 @@ +repos: + - repo: https://github.com/pre-commit/pre-commit-hooks + rev: v4.3.0 + hooks: + - id: check-yaml + - id: end-of-file-fixer + - id: trailing-whitespace + - id: check-added-large-files + args: + - "--maxkb=50000" + - repo: local + hooks: + - id: check-illegal-windows-names + name: check illegal windows names + language: fail + entry: Illegal windows filenames detected + verbose: True + files: ':|<|>|;|,|\*|\?|\\' diff --git a/content/about.md b/content/about.md index 87dcfa5..7a766d7 100644 --- a/content/about.md +++ b/content/about.md @@ -10,4 +10,4 @@ In particular, we are inspired by recent results in theoretical neuroscience (th Our mission is to translate these ideas to *synthetic* agents that automate the design of signal processing and control systems through unsupervised environmental interactions. Applications range from situated design of multimedia applications (e.g., personalized hearing aid algorithms) to real-time decision making for robots and drones. Please view a short introduction video below for more info. -{{}} \ No newline at end of file +{{}} diff --git a/content/alumni/_index.md b/content/alumni/_index.md index cb7c3b6..b17b217 100644 --- a/content/alumni/_index.md +++ b/content/alumni/_index.md @@ -1,3 +1,3 @@ +++ title = "Alumni" -+++ \ No newline at end of file ++++ diff --git a/content/alumni/anouk.md b/content/alumni/anouk.md index 6b31a97..527e59c 100644 --- a/content/alumni/anouk.md +++ b/content/alumni/anouk.md @@ -30,4 +30,4 @@ external_link = "https://research.tue.nl/en/persons/anouk-van-diepen" +++ -Anouk van Diepen received her M.Sc. degree in Electrical Engineering from the TU Eindhoven in 2017. \ No newline at end of file +Anouk van Diepen received her M.Sc. degree in Electrical Engineering from the TU Eindhoven in 2017. diff --git a/content/alumni/burak.md b/content/alumni/burak.md index 4b9d826..ad4b8f3 100644 --- a/content/alumni/burak.md +++ b/content/alumni/burak.md @@ -29,4 +29,4 @@ external_link = "https://www.linkedin.com/in/burak-ergul-/" role = "Former MSc student" +++ -Burak Ergul was an MSc student in the SPS group of Electrical Engineering department in Eindhoven University of Technology. His research focused on real-world applications of Active Inference on embedded devices. \ No newline at end of file +Burak Ergul was an MSc student in the SPS group of Electrical Engineering department in Eindhoven University of Technology. His research focused on real-world applications of Active Inference on embedded devices. diff --git a/content/alumni/eric.md b/content/alumni/eric.md index c4fa387..b2958b5 100644 --- a/content/alumni/eric.md +++ b/content/alumni/eric.md @@ -24,4 +24,4 @@ is_former_member = true role = "Former PhD candidate" +++ -Quan (Eric) Nguyen received a M.Sc. in Computer Science (with a specialization in Algorithms, Data Analytics and Machine Learning; and a minor in Statistics, both with distinction) from the University of Helsinki in 2015. He left the group in 2018. \ No newline at end of file +Quan (Eric) Nguyen received a M.Sc. in Computer Science (with a specialization in Algorithms, Data Analytics and Machine Learning; and a minor in Statistics, both with distinction) from the University of Helsinki in 2015. He left the group in 2018. diff --git a/content/alumni/jim.md b/content/alumni/jim.md index e131d9d..96d96c4 100644 --- a/content/alumni/jim.md +++ b/content/alumni/jim.md @@ -1,38 +1,38 @@ -+++ -date = "2021-09-01T15:52:22+02:00" -id = "jim" ++++ +date = "2021-09-01T15:52:22+02:00" +id = "jim" interests = ["Signal Processing", "Bayesian Machine Learning", "Speech Recognition"] -name = " Jim Beckers" -portrait = "portraits/Jim.jpg" -short_bio = "I am a MSc student at the SPS group at Eindhoven University of Technology." -title = "Jim Beckers" +name = " Jim Beckers" +portrait = "portraits/Jim.jpg" +short_bio = "I am a MSc student at the SPS group at Eindhoven University of Technology." +title = "Jim Beckers" sort_position = 6 external_link = "https://www.linkedin.com/in/jim-beckers/" -[[social]] - icon = "envelope" - icon_pack = "fa" +[[social]] + icon = "envelope" + icon_pack = "fa" link = "mailto:j.j.p.beckers@student.tue.nl" -[[social]] - icon = "github" - icon_pack = "fa" +[[social]] + icon = "github" + icon_pack = "fa" link = "//github.com/JimBeckers/" -[[social]] - icon = "linkedin" - icon_pack = "fa" +[[social]] + icon = "linkedin" + icon_pack = "fa" link = "https://www.linkedin.com/in/jim-beckers/" -[[education]] - course = "BSc. in Electrical Engineering" - institution = 'Eindhoven University of Technology' +[[education]] + course = "BSc. in Electrical Engineering" + institution = 'Eindhoven University of Technology' year = 2019 -[[organizations]] - name = "TU Eindhoven" +[[organizations]] + name = "TU Eindhoven" role = "MSc student" +++ -Jim Beckers was a MSc student in the SPS group of the Electrical Engineering department at Eindhoven University of Technology. His research focused on implementing the Active Listening paper by Karl Friston. \ No newline at end of file +Jim Beckers was a MSc student in the SPS group of the Electrical Engineering department at Eindhoven University of Technology. His research focused on implementing the Active Listening paper by Karl Friston. diff --git a/content/alumni/marco.md b/content/alumni/marco.md index 1d30f09..8e1ef3e 100644 --- a/content/alumni/marco.md +++ b/content/alumni/marco.md @@ -31,4 +31,3 @@ external_link = "https://www.linkedin.com/in/marco-cox-a455a01b/" +++ Marco Cox was a PhD candidate in the Electrical Engineering department of [TU Eindhoven](http://www.tue.nl), working in the [BIASlab](http://biaslab.github.io) team. His research focused on Bayesian machine learning techniques for optimization and signal processing. He is also interested in information theory, computer architecture, and combinatorics. Marco received the M.Sc. degree in electrical engineering from TU Eindhoven in 2014. - diff --git a/content/alumni/mark.md b/content/alumni/mark.md index 278f079..70ba115 100644 --- a/content/alumni/mark.md +++ b/content/alumni/mark.md @@ -1,27 +1,27 @@ -+++ -date = "2020-08-27T11:58:58+02:00" -id = "mark" -interests = ["Signal Processing", "Artificial Intelligence"] -name = "Mark Legters" -portrait = "portraits/mark.jpg" -short_bio = "I am a MSc student at BIASlab, conducting my graduation project in collaboration with [Sorama](http://sorama.eu)." -title = "Mark Legters" ++++ +date = "2020-08-27T11:58:58+02:00" +id = "mark" +interests = ["Signal Processing", "Artificial Intelligence"] +name = "Mark Legters" +portrait = "portraits/mark.jpg" +short_bio = "I am a MSc student at BIASlab, conducting my graduation project in collaboration with [Sorama](http://sorama.eu)." +title = "Mark Legters" sort_position = 7 -external_link = "https://www.linkedin.com/in/marklegters/" +external_link = "https://www.linkedin.com/in/marklegters/" -[[social]] - icon = "linkedin" - icon_pack = "fa" +[[social]] + icon = "linkedin" + icon_pack = "fa" link = "//linkedin.com/in/marklegters/" -[[education]] - course = "BSc in Electrical Engineering" - institution = 'Eindhoven University of Technology' +[[education]] + course = "BSc in Electrical Engineering" + institution = 'Eindhoven University of Technology' year = 2017 -[[organizations]] - name = "TU Eindhoven" +[[organizations]] + name = "TU Eindhoven" role = "MSc student" -+++ -Mark Legters was an MSc student in the [SPS group](https://www.tue.nl/en/research/research-groups/signal-processing-systems/) of the Electrical Engineering department at the [Eindhoven University of Technology](http://tue.nl). He was conducting his MSc graduation project at [BIASlab](http://biaslab.github.io), in collaboration with [Sorama](http://sorama.eu). His research focused on In-situ Learning of Acoustic Event Detection using Nature-inspired Artificial Intelligence. \ No newline at end of file ++++ +Mark Legters was an MSc student in the [SPS group](https://www.tue.nl/en/research/research-groups/signal-processing-systems/) of the Electrical Engineering department at the [Eindhoven University of Technology](http://tue.nl). He was conducting his MSc graduation project at [BIASlab](http://biaslab.github.io), in collaboration with [Sorama](http://sorama.eu). His research focused on In-situ Learning of Acoustic Event Detection using Nature-inspired Artificial Intelligence. diff --git a/content/alumni/patrick.md b/content/alumni/patrick.md index 1d0f90d..72e01f7 100644 --- a/content/alumni/patrick.md +++ b/content/alumni/patrick.md @@ -22,4 +22,4 @@ external_link = "https://research.tue.nl/en/persons/patrick-wijnings" name = "TU Eindhoven" role = "PhD candidate" +++ -Patrick Wijnings got his MSc in Electrical Engineering at [Eindhoven University of Technology](http://tue.nl) in 2015. Since 2011 he has been working (part time) at [Sorama](http://sorama.eu) as System Engineer, where he helped develop the hardware and real-time signal processing core used in [the largest microphone array in the world](https://www.guinnessworldrecords.com/world-records/largest-microphone-array/). Since July 2017 he was pursuing his PhD at [Eindhoven University of Technology](http://tue.nl) in collaboration with [BIASlab](http://biaslab.github.io) and [Sorama](http://sorama.eu). His research interests are Bayesian inverse acoustics for large microphone arrays, and the accompanying problem of approximate inference for high-dimensional models. \ No newline at end of file +Patrick Wijnings got his MSc in Electrical Engineering at [Eindhoven University of Technology](http://tue.nl) in 2015. Since 2011 he has been working (part time) at [Sorama](http://sorama.eu) as System Engineer, where he helped develop the hardware and real-time signal processing core used in [the largest microphone array in the world](https://www.guinnessworldrecords.com/world-records/largest-microphone-array/). Since July 2017 he was pursuing his PhD at [Eindhoven University of Technology](http://tue.nl) in collaboration with [BIASlab](http://biaslab.github.io) and [Sorama](http://sorama.eu). His research interests are Bayesian inverse acoustics for large microphone arrays, and the accompanying problem of approximate inference for high-dimensional models. diff --git a/content/alumni/philip.md b/content/alumni/philip.md index 5ce82a1..2a3801a 100644 --- a/content/alumni/philip.md +++ b/content/alumni/philip.md @@ -1,5 +1,5 @@ -+++ -date = "2021-09-10T09:00:00+02:00" ++++ +date = "2021-09-10T09:00:00+02:00" id = "philip" interests = ["Machine Learning", "Signal Processing"] name = "Philip Spannring" @@ -9,27 +9,27 @@ title = "Philip Spannring" sort_position = 7 external_link = "https://www.linkedin.com/in/philip-spannring/" -[[social]] +[[social]] icon = "envelope" icon_pack = "fa" link = "mailto:p.spannring@student.tue.nl" -[[social]] +[[social]] icon = "github" icon_pack = "fa" link = "//github.com/philipspannring" -[[social]] +[[social]] icon = "linkedin" icon_pack = "fa" link = "https://www.linkedin.com/in/philip-spannring/" -[[education]] +[[education]] course = "BSc. in Electrical Engineering" institution = 'Eindhoven University of Technology' year = 2019 -[[organizations]] +[[organizations]] name = "TU Eindhoven" role = "MSc student" +++ diff --git a/content/alumni/thijn.md b/content/alumni/thijn.md index 63c1a19..56ae59e 100644 --- a/content/alumni/thijn.md +++ b/content/alumni/thijn.md @@ -1,37 +1,37 @@ -+++ -date = "2020-09-01T15:52:22+02:00" -id = "thijn" -interests = ["Machine Learning", "Generative modeling", "Sports and Statistics", "Markets and Trading"] -name = "Thijn Hermsen" -portrait = "portraits/Thijn.jpg" -short_bio = "I am a MSc student at the SPS group at Eindhoven University of Technology working on Generative modeling in football player's decisions." -title = "Thijn Hermsen" ++++ +date = "2020-09-01T15:52:22+02:00" +id = "thijn" +interests = ["Machine Learning", "Generative modeling", "Sports and Statistics", "Markets and Trading"] +name = "Thijn Hermsen" +portrait = "portraits/Thijn.jpg" +short_bio = "I am a MSc student at the SPS group at Eindhoven University of Technology working on Generative modeling in football player's decisions." +title = "Thijn Hermsen" sort_position = 6 external_link = "https://www.linkedin.com/in/thijn-hermsen-electrical-engineering/" -[[social]] - icon = "envelope" - icon_pack = "fa" +[[social]] + icon = "envelope" + icon_pack = "fa" link = "mailto:t.w.t.hermsen@student.tue.nl" -[[social]] - icon = "github" - icon_pack = "fa" +[[social]] + icon = "github" + icon_pack = "fa" link = "//github.com/ThijnHermsen/" -[[social]] - icon = "linkedin" - icon_pack = "fa" +[[social]] + icon = "linkedin" + icon_pack = "fa" link = "https://www.linkedin.com/in/thijn-hermsen-electrical-engineering/" -[[education]] - course = "BSc. in Electrical Engineering" - institution = 'Eindhoven University of Technology' +[[education]] + course = "BSc. in Electrical Engineering" + institution = 'Eindhoven University of Technology' year = 2018 -[[organizations]] - name = "TU Eindhoven" +[[organizations]] + name = "TU Eindhoven" role = "MSc student" +++ -My name is Thijn Hermsen and I was a MSc student in the Signal Processing Systems group of the Electrical Engineering department at Eindhoven University of Technology. My research focused on Generative modeling on football player's decisions and further interests extend towards statistics and market strategies. \ No newline at end of file +My name is Thijn Hermsen and I was a MSc student in the Signal Processing Systems group of the Electrical Engineering department at Eindhoven University of Technology. My research focused on Generative modeling on football player's decisions and further interests extend towards statistics and market strategies. diff --git a/content/member/_index.md b/content/member/_index.md index 992684e..0356288 100644 --- a/content/member/_index.md +++ b/content/member/_index.md @@ -1,4 +1,4 @@ +++ title = "BIASlab team" url="/team" -+++ \ No newline at end of file ++++ diff --git a/content/member/bart.md b/content/member/bart.md index b82b3e9..f147f76 100644 --- a/content/member/bart.md +++ b/content/member/bart.md @@ -12,7 +12,7 @@ sort_position = 4.2 icon = "envelope" icon_pack = "fa" link = "mailto:b.v.erp@tue.nl" - + [[social]] icon = "linkedin" icon_pack = "fa" @@ -39,7 +39,7 @@ sort_position = 4.2 +++ -Bart has obtained his Bachelor and Master degrees in Electrical Engineering at the Eindhoven University of Technology in 2018 and 2020, respectively. -Currently he is combining his passion for research and education as a PhD candidate - Teaching Assistant at the Eindhoven University of Technology in the BIASlab research group. -His research focuses on the design of new signal processing algorithms for audio and speech. -These algorithms are inspired by conventional approaches and are extended to allow for on-the-spot user customization. \ No newline at end of file +Bart has obtained his Bachelor and Master degrees in Electrical Engineering at the Eindhoven University of Technology in 2018 and 2020, respectively. +Currently he is combining his passion for research and education as a PhD candidate - Teaching Assistant at the Eindhoven University of Technology in the BIASlab research group. +His research focuses on the design of new signal processing algorithms for audio and speech. +These algorithms are inspired by conventional approaches and are extended to allow for on-the-spot user customization. diff --git a/content/member/bert.md b/content/member/bert.md index a2c7acc..c23378c 100644 --- a/content/member/bert.md +++ b/content/member/bert.md @@ -40,4 +40,4 @@ cv_link = "http://bertdv.github.io/cv/bdevries_cv.pdf" +++ Bert de Vries received MSc (1986) and PhD (1991) degrees in Electrical Engineering from [Eindhoven University of Technology](http://tue.nl) (TU/e) and the [University of Florida](http://ufl.edu), respectively. From 1992 until 1999 he worked at [Sarnoff Research Center](https://www.sri.com/) in Princeton (NJ), where he contributed to research projects over a wide range of signal and image processing topics such as word spotting, financial market prediction, and breast cancer detection from mammograms. Since April 1999 he has been employed in the hearing aids industry (currently at [GN Hearing](http://gnhearing.com)), both in research and managerial roles. Since January 2012 he is also a full professor at the Signal Processing Systems Group at TU/e, where he teaches a [course on Bayesian machine learning](http://bmlip.nl) to graduate electrical engineering students. At TU/e he directs the [BIASlab](http://biaslab.github.io) research team of graduate students with whom he conducts [research on transferring - a Bayesian brain theory (the Free Energy Principle) to practical engineering](https://youtu.be/QYbcm6G_wsk). \ No newline at end of file + a Bayesian brain theory (the Free Energy Principle) to practical engineering](https://youtu.be/QYbcm6G_wsk). diff --git a/content/member/bvdmitri.md b/content/member/bvdmitri.md index 76183cc..e2b6456 100644 --- a/content/member/bvdmitri.md +++ b/content/member/bvdmitri.md @@ -45,6 +45,3 @@ sort_position = 4.1 role = "PhD candidate" +++ My research interests lie in the fields of computers science, machine learning and probabilistic programming. Currently I am a PhD candidate in the SPS group of Electrical Engineering department in Eindhoven University of Technology. I'm working on a high-performant implementation of message passing-based Bayesian inference package in the Julia programming language. My research project focuses on Signal Processing and Active inference applications, but is also aimed to expand the scope of possible applications for message passing in general. - - - diff --git a/content/member/hoang.md b/content/member/hoang.md index 131fc56..4531e7e 100644 --- a/content/member/hoang.md +++ b/content/member/hoang.md @@ -1,37 +1,37 @@ -+++ ++++ date = "2016-07-12T15:52:22+02:00" -id = "hoang" -interests = ["Differentiable Programming for Speech and Audio Processing", "Bayesian Machine Learning", "Signal Processing"] -name = "Hoang Minh Huu Nguyen" -portrait = "portraits/Hoang.jpeg" -short_bio = "I am a PhD candidate at the Signal Processing Systems group in TU Eindhoven working on Bayesian Machine Learning." -title = "Hoang M.H. Nguyen" +id = "hoang" +interests = ["Differentiable Programming for Speech and Audio Processing", "Bayesian Machine Learning", "Signal Processing"] +name = "Hoang Minh Huu Nguyen" +portrait = "portraits/Hoang.jpeg" +short_bio = "I am a PhD candidate at the Signal Processing Systems group in TU Eindhoven working on Bayesian Machine Learning." +title = "Hoang M.H. Nguyen" sort_position = 4.3 -[[social]] -icon = "envelope" -icon_pack = "fa" +[[social]] +icon = "envelope" +icon_pack = "fa" link = "mailto:m.h.n.hoang@tue.nl" -[[social]] -icon = "github" -icon_pack = "fa" -link = "//github.com/HoangMHNguyen/" +[[social]] +icon = "github" +icon_pack = "fa" +link = "//github.com/HoangMHNguyen/" -[[education]] -course = "MSc in Electrical Engineering" -institution = 'Eindhoven University of Technology' +[[education]] +course = "MSc in Electrical Engineering" +institution = 'Eindhoven University of Technology' year = 2021 -[[education]] -course = "BSc in Electrical Engineering" -institution = 'Vietnam National University HCMC-International University' +[[education]] +course = "BSc in Electrical Engineering" +institution = 'Vietnam National University HCMC-International University' year = 2018 -[[organizations]] -name = "TU Eindhoven" +[[organizations]] +name = "TU Eindhoven" role = "PhD candidate" -+++ -Hoang Minh Huu Nguyen is a PhD candidate in the SPS group of the Electrical Engineering department at the Eindhoven University of Technology. His research focuses on developing usable synthetic active inference agents through the development of scalable real-time Bayesian inference.. \ No newline at end of file ++++ +Hoang Minh Huu Nguyen is a PhD candidate in the SPS group of the Electrical Engineering department at the Eindhoven University of Technology. His research focuses on developing usable synthetic active inference agents through the development of scalable real-time Bayesian inference.. diff --git a/content/member/ismail.md b/content/member/ismail.md index b666aaf..1eb56cb 100644 --- a/content/member/ismail.md +++ b/content/member/ismail.md @@ -40,4 +40,4 @@ sort_position = 3.0 role = "Postdoctoral fellow" +++ -Ismail Senoz received his M.Sc. degree in electrical engineering from the Eindhoven University of Technology in 2017 and his Ph.D. in machine learning from the Eindhoven University of Technology in 2022. Currently, he works as a postdoctoral fellow at BIASlab at the Eindhoven University of Technology. His research focuses on approximate message passing algorithms. +Ismail Senoz received his M.Sc. degree in electrical engineering from the Eindhoven University of Technology in 2017 and his Ph.D. in machine learning from the Eindhoven University of Technology in 2022. Currently, he works as a postdoctoral fellow at BIASlab at the Eindhoven University of Technology. His research focuses on approximate message passing algorithms. diff --git a/content/member/jia.md b/content/member/jia.md index 67bad01..0377ad8 100644 --- a/content/member/jia.md +++ b/content/member/jia.md @@ -12,7 +12,7 @@ sort_position = 6.0 icon = "envelope" icon_pack = "fa" link = "c.jia@tue.nl" - + [[social]] icon = "github" icon_pack = "fa" @@ -22,7 +22,7 @@ sort_position = 6.0 course = "MSc in Mathematics and applied mathematics" institution = 'Central South University for Nationalities' year = 2016 - + [[education]] course = "BSc in Mathematics" institution = 'Wuhan University of Technology' @@ -36,5 +36,5 @@ sort_position = 6.0 +++ Chengfeng Jia has received a BSc degree in mathematics and applied mathematics from the Central South University for Nationalities, China, in 2016, and an MSs degree in mathematics from the Wuhan University of Technology (WHUT), China, in 2019. -He is currently pursuing a PhD degree in the school of Navigation at Wuhan University of Technology, and he is the guest PhD in Electrical Engineering at the Eindhoven University of Technology in the BIASlab research group. -His research interests focus on Bayesian Machine Learning, Intelligent Navigation, and Intelligent transportation systems. \ No newline at end of file +He is currently pursuing a PhD degree in the school of Navigation at Wuhan University of Technology, and he is the guest PhD in Electrical Engineering at the Eindhoven University of Technology in the BIASlab research group. +His research interests focus on Bayesian Machine Learning, Intelligent Navigation, and Intelligent transportation systems. diff --git a/content/member/magnus.md b/content/member/magnus.md index 5e1e9f8..e8c20e0 100644 --- a/content/member/magnus.md +++ b/content/member/magnus.md @@ -4,7 +4,7 @@ id = "magnus" interests = ["Computational Neuroscience" ,"Bayesian Machine Learning", "Reinforcement Learning"] name = "Magnus Tønder Koudahl" portrait = "portraits/magnus.png" -short_bio = "I am a PhD candidate at the Signal Processing Systems group in TU Eindhoven" +short_bio = "I am a PhD candidate at the Signal Processing Systems group in TU Eindhoven" sort_position = 4.0 title = "Magnus Koudahl" @@ -21,13 +21,13 @@ title = "Magnus Koudahl" [[education]] course = "M.Sc in IT & Cognition" institution = 'University of Copenhagen' - year = 2017 + year = 2017 [[education]] course = "B.Sc in Psychology" institution = 'University of Copenhagen' - year = 2015 + year = 2015 [[organizations]] name = "TU Eindhoven" diff --git a/content/member/sepideh.md b/content/member/sepideh.md index 4a9a5bc..a799acc 100644 --- a/content/member/sepideh.md +++ b/content/member/sepideh.md @@ -27,7 +27,7 @@ sort_position = 4.6 course = "BSc in Computer Science" institution = 'University of Isfahan' year = 2017 - + [[organizations]] name = "TU Eindhoven" role = "PhD candidate" diff --git a/content/member/wouter.md b/content/member/wouter.md index e221b83..e1e676b 100644 --- a/content/member/wouter.md +++ b/content/member/wouter.md @@ -46,4 +46,3 @@ cv_link = "" +++ Wouter Kouw has a dual background in neuroscience and computer science, which he utilizes to analyze, design and develop probabilistic machine learning systems based on models of information processing in the brain. His research focuses on variational Bayesian inference algorithms for learning dynamics in mobile robots. Formerly, he was a visiting scholar at Cornell University (USA) and a Niels Stensen Fellow at Copenhagen University (Denmark). - diff --git a/content/member/woutern.md b/content/member/woutern.md index 5ad979b..3d5a578 100644 --- a/content/member/woutern.md +++ b/content/member/woutern.md @@ -42,4 +42,4 @@ cv_link = "" +++ -Wouter Nuijten is a PhD candidate in the SPS group at Eindhoven University of Technology's Electrical Engineering department. His research focuses on the study of nested Active Inference and the emergence of intelligence within complex hierarchical systems. The goal of this research is to design intelligent agents that minimize variational free energy in a control setting. \ No newline at end of file +Wouter Nuijten is a PhD candidate in the SPS group at Eindhoven University of Technology's Electrical Engineering department. His research focuses on the study of nested Active Inference and the emergence of intelligence within complex hierarchical systems. The goal of this research is to design intelligent agents that minimize variational free energy in a control setting. diff --git a/content/open-projects/_index.md b/content/open-projects/_index.md index 027562c..bd8fd03 100644 --- a/content/open-projects/_index.md +++ b/content/open-projects/_index.md @@ -1,3 +1,3 @@ +++ title="Open projects" -+++ \ No newline at end of file ++++ diff --git a/content/open-projects/msc-cocktail-party.md b/content/open-projects/msc-cocktail-party.md index 2c035d1..c5a4dc9 100644 --- a/content/open-projects/msc-cocktail-party.md +++ b/content/open-projects/msc-cocktail-party.md @@ -14,7 +14,7 @@ You are challenged to design an agent that learns to solve the cocktail party problem through on-the-spot interactions with a (human) listener. The cocktail party problem refers to the issue of not being able to understand your conversation partner in the presence -of many simultaneously competing voices (Fig.1). +of many simultaneously competing voices (Fig.1). {{< figure src="/img/proposals/cocktail-party.jpeg" title="Sound signals from multiple conversations mix at a cocktail party." class="center" height="100px" >}} @@ -61,4 +61,4 @@ computational neuroscience and signal processing. - The student should prepare update meetings, preferably with derivations or visualizations in interactive notebooks. ## Timeline -The project is available from September 2017 onwards. The total duration will be 32 weeks. Halfway through, there will be a "midterm" evaluation where the student must report on their activities and indicate how they will proceed for the remainder of the project. At the end, the student will write a thesis summarizing their work, their findings and possible future steps. The thesis will be presented in an official "defense" ceremony and a committee of experts will grade the student's work. \ No newline at end of file +The project is available from September 2017 onwards. The total duration will be 32 weeks. Halfway through, there will be a "midterm" evaluation where the student must report on their activities and indicate how they will proceed for the remainder of the project. At the end, the student will write a thesis summarizing their work, their findings and possible future steps. The thesis will be presented in an official "defense" ceremony and a committee of experts will grade the student's work. diff --git a/content/open-projects/msc-multi-agent-interaction.md b/content/open-projects/msc-multi-agent-interaction.md index 0157fc3..9f086a3 100644 --- a/content/open-projects/msc-multi-agent-interaction.md +++ b/content/open-projects/msc-multi-agent-interaction.md @@ -60,4 +60,4 @@ For more information go to the project page at [TU/e Master Marketplace](https:/ [3]: L. Pio-Lopez, A. Nizard, K. J. Friston, and G. Pezzulo. Active inference and robot control: a case study. *Journal of The Royal Society Interface, 13(122):20160616, 2016*. -[4]: T. W. van de Laar and B. de Vries, Simulating active inference processes by message passing, *Frontiers in Robotics and AI, vol. 6, p. 20, 2019*. \ No newline at end of file +[4]: T. W. van de Laar and B. de Vries, Simulating active inference processes by message passing, *Frontiers in Robotics and AI, vol. 6, p. 20, 2019*. diff --git a/content/open-projects/msc-nonlinear-system-identification-through-bayesian-machine-learning.md b/content/open-projects/msc-nonlinear-system-identification-through-bayesian-machine-learning.md index 1df2703..48df102 100644 --- a/content/open-projects/msc-nonlinear-system-identification-through-bayesian-machine-learning.md +++ b/content/open-projects/msc-nonlinear-system-identification-through-bayesian-machine-learning.md @@ -83,4 +83,4 @@ dynamics as fast as possible. For more information go to the project page at [TU/e Master Marketplace](https://master.ele.tue.nl/). ## References -[1]: A. Janot, M. Gautier, and M. Brunot. Data set and reference models of EMPS. In *2019 Workshop on Nonlinear System Identification Benchmarks, Eindhoven, The Netherlands, April 10-12, 2019*. \ No newline at end of file +[1]: A. Janot, M. Gautier, and M. Brunot. Data set and reference models of EMPS. In *2019 Workshop on Nonlinear System Identification Benchmarks, Eindhoven, The Netherlands, April 10-12, 2019*. diff --git a/content/open-projects/msc-parallel-reactive-computing.md b/content/open-projects/msc-parallel-reactive-computing.md index 33d2369..5c7b880 100644 --- a/content/open-projects/msc-parallel-reactive-computing.md +++ b/content/open-projects/msc-parallel-reactive-computing.md @@ -34,4 +34,4 @@ Intelligent agents process information through Bayesian inference. In our lab, w - All developed code should be accessible (e.g., on BIASlab's Github organization). ## Timeline -The project is available from September 2022 onwards. The total duration will be 32 weeks. Halfway through, there will be a "midterm" evaluation where the student must report on their activities and indicate how they will proceed for the remainder of the project. At the end, the student will write a thesis summarizing their work, their findings and possible future steps. The thesis will be presented in an official "defense" ceremony and a committee of experts will grade the student's work. \ No newline at end of file +The project is available from September 2022 onwards. The total duration will be 32 weeks. Halfway through, there will be a "midterm" evaluation where the student must report on their activities and indicate how they will proceed for the remainder of the project. At the end, the student will write a thesis summarizing their work, their findings and possible future steps. The thesis will be presented in an official "defense" ceremony and a committee of experts will grade the student's work. diff --git a/content/open-projects/msc-quadruped-locomotion.md b/content/open-projects/msc-quadruped-locomotion.md index 7e5f04a..22c5238 100644 --- a/content/open-projects/msc-quadruped-locomotion.md +++ b/content/open-projects/msc-quadruped-locomotion.md @@ -8,11 +8,11 @@ vacancy_id = "quadruped" +++ -The goal is to develop an intelligent autonomous system (agent) for a quadrupedal robot (see example in Figure 1). The agent must learn to walk: it will not be given an accurate model of its legs' kinematics but will have to gradually build a locomotion model from interacting with its environment. You will use _Active Inference_ (AIF), a probabilistic machine learning framework from the computational neuroscience community, to design and train the agent. +The goal is to develop an intelligent autonomous system (agent) for a quadrupedal robot (see example in Figure 1). The agent must learn to walk: it will not be given an accurate model of its legs' kinematics but will have to gradually build a locomotion model from interacting with its environment. You will use _Active Inference_ (AIF), a probabilistic machine learning framework from the computational neuroscience community, to design and train the agent. {{< figure src="/img/proposals/SpotMicroAI_complete_1.jpg" title="A physical Spotmicro build. Figure courtesy of https://spotmicroai.readthedocs.io." class="center" height="100px" >}} -Please note that this is a software project; it does not involve hardware. You will write code to control a simulated robot that interacts with a simulated environment (see Figure 2), generated using the open-source physics engine [Bullet](https://pybullet.org/wordpress/). The challenge will be to adapt existing AIF agents to control the robotic system. +Please note that this is a software project; it does not involve hardware. You will write code to control a simulated robot that interacts with a simulated environment (see Figure 2), generated using the open-source physics engine [Bullet](https://pybullet.org/wordpress/). The challenge will be to adapt existing AIF agents to control the robotic system. {{< figure src="/img/proposals/spot-mini-mini.gif" title="A simulated Spotmicro in a physics simulation. Figure courtesy of https://spotmicroai.readthedocs.io." class="center" height="100px" >}} @@ -26,16 +26,16 @@ This project does not involve a company. You will be working in the Bayesian Int - Developed software will be open source, accessible through [BIASlab's github](https://github.com/biaslab/) organization. It should be legible / usable for future students. ## Student task description -You will initially be spending time familiarizing yourself with the tools and techniques that BIASlab develops ([RxInfer.jl](https://github.com/biaslab/RxInfer.jl)). Once familiar, you will write your own active inference agent based off of existing AIF agent implementations within BIASlab. Note that you'll be supported by BIASlab researchers that are working on other robot locomotion projects. +You will initially be spending time familiarizing yourself with the tools and techniques that BIASlab develops ([RxInfer.jl](https://github.com/biaslab/RxInfer.jl)). Once familiar, you will write your own active inference agent based off of existing AIF agent implementations within BIASlab. Note that you'll be supported by BIASlab researchers that are working on other robot locomotion projects. ### Concrete tasks - Review literature on AIF agents for robotics. - Familiarize yourself with the challenges of quadrupedal locomotion. - Learn to use the probabilistic machine learning toolbox RxInfer.jl. - Familiarize yourself with the simulation environment Bullet. -- Develop an active inference agent for a quadrupedal robot system. -- Reflect on what has been achieved and discuss with BIASlab's researchers. +- Develop an active inference agent for a quadrupedal robot system. +- Reflect on what has been achieved and discuss with BIASlab's researchers. - Write a report detailing your agent's properties and behaviour. ## Timeline -The project is available from September 2022 onwards. The total duration will be 32 weeks. Halfway through, there will be a "midterm" evaluation where the student must report on their activities and indicate how they will proceed for the remainder of the project. At the end, the student will write a thesis summarizing their work, their findings and possible future steps. The thesis will be presented in an official "defense" ceremony and a committee of experts will grade the student's work. \ No newline at end of file +The project is available from September 2022 onwards. The total duration will be 32 weeks. Halfway through, there will be a "midterm" evaluation where the student must report on their activities and indicate how they will proceed for the remainder of the project. At the end, the student will write a thesis summarizing their work, their findings and possible future steps. The thesis will be presented in an official "defense" ceremony and a committee of experts will grade the student's work. diff --git a/content/post/approach.md b/content/post/approach.md index 00654c4..6cf3bf3 100644 --- a/content/post/approach.md +++ b/content/post/approach.md @@ -30,12 +30,12 @@ In order to scale down a design iteration from years to seconds, we cannot affor HA design with IA -The task of the automated designer, which in the technical literature is called an **intelligent agent** (IA), is to propose an _interesting_ alternative algorithm each time when the patient is not happy, under in-situ conditions. This is a monumental task that involves learning from experiences and rational decision making under uncertainty. We take a fully Bayesian (= probabilistic) approach to designing such intelligent agents (see also our [mission](/mission) page). +The task of the automated designer, which in the technical literature is called an **intelligent agent** (IA), is to propose an _interesting_ alternative algorithm each time when the patient is not happy, under in-situ conditions. This is a monumental task that involves learning from experiences and rational decision making under uncertainty. We take a fully Bayesian (= probabilistic) approach to designing such intelligent agents (see also our [mission](/mission) page). ### The broader picture Next to hearing aid design, Bayesian intelligent agents may have applications to solving problems whenever we don't have a clean problem description. Most interesting problems are of this nature. In our team, the focus lies on applications to hearing aids and hearables for health and fitness monitoring and coaching, but we are also interested in other wearable smart computing applications. -Finally, the idea of focusing on fast iterations as a fundamental _design principle_ has permeated various related disciplines. In the context of software design, Sandi Metz summarizes the idea as follows (pg.16 in [Practical Object-Oriented Design in Ruby](http://poodr.com), 2012): +Finally, the idea of focusing on fast iterations as a fundamental _design principle_ has permeated various related disciplines. In the context of software design, Sandi Metz summarizes the idea as follows (pg.16 in [Practical Object-Oriented Design in Ruby](http://poodr.com), 2012): > Design is more the art of preserving changeability than it is the art of achieving perfection. diff --git a/content/project/forneylab.md b/content/project/forneylab.md index 61f437a..7de316c 100644 --- a/content/project/forneylab.md +++ b/content/project/forneylab.md @@ -67,4 +67,4 @@ ForneyLab is especially potent when applied to time-series data. In this section [4] Loeliger, H-A. "An introduction to factor graphs." _IEEE Signal Processing Magazine 21.1_ (2004): 28-41. -[5] Cox, M., van de Laar, T., de Vries, B. "A Factor Graph Approach to Automated Design of Bayesian Signal Processing Algorithms", International Journal of Approximate Reasoning (2019), 10.1016/j.ijar.2018.11.002. \ No newline at end of file +[5] Cox, M., van de Laar, T., de Vries, B. "A Factor Graph Approach to Automated Design of Bayesian Signal Processing Algorithms", International Journal of Approximate Reasoning (2019), 10.1016/j.ijar.2018.11.002. diff --git a/content/project/learning-where-to-park-by-active-inference.md b/content/project/learning-where-to-park-by-active-inference.md index 6903e1c..757ccd6 100644 --- a/content/project/learning-where-to-park-by-active-inference.md +++ b/content/project/learning-where-to-park-by-active-inference.md @@ -16,32 +16,32 @@ sort_position = 1 +++ ## Problem Statement -The idea of autonomously operating ("self-learning") information processing agents has recently gained traction in the AI research community. Depending on context, the development of these agents involves some hard challenges including the need for the agents to be capable of learning specific goals in dynamic real-world. +The idea of autonomously operating ("self-learning") information processing agents has recently gained traction in the AI research community. Depending on context, the development of these agents involves some hard challenges including the need for the agents to be capable of learning specific goals in dynamic real-world. In this project we investigated a novel solution approach to the design of autonomous agents. We recognize that any "intelligent" autonomously operating agent needs to be minimally capable of realizing three tasks: 1. Perception: online tracking of the state of the world. 1. Learning: making updates to its world model in case the agent's model poorly predicts real-world dynamics. -1. Decision making: agents should be able to act onto ("control") the world in order to make an impact on the future dynamics of the world. +1. Decision making: agents should be able to act onto ("control") the world in order to make an impact on the future dynamics of the world. There exists a powerful computational theory for how _biological_ agents accomplish the aforementioned task palette. This theory, which is called _Active Inference_, relies on formulating all tasks (perception, learning and control) as (automatable) inference tasks in a (biased) generative model of agent's sensory inputs [[3](#references)]. - -In order to assess the feasibility and capabilities of active inference as a framework for _synthetic_ agents in a real-world setting, we developed a ground-based robot that needs to learn to navigate to an undisclosed parking location. The robot can only learn where to park through situated interactions with a human observer who is aware of the target location. + +In order to assess the feasibility and capabilities of active inference as a framework for _synthetic_ agents in a real-world setting, we developed a ground-based robot that needs to learn to navigate to an undisclosed parking location. The robot can only learn where to park through situated interactions with a human observer who is aware of the target location. ## Methods and Solution Proposal -Figure 1 shows the robot that was used in this project. The Boe-Bot robot by Parallax comes equipped with wheels and continuous rotation servomotors [[5](#references)]. The computational backbone relies on an Arduino Uno to gather sensor signals and the Active Inference agent was coded in the Julia programming language on a Raspberry Pi4-based Linux environment. +Figure 1 shows the robot that was used in this project. The Boe-Bot robot by Parallax comes equipped with wheels and continuous rotation servomotors [[5](#references)]. The computational backbone relies on an Arduino Uno to gather sensor signals and the Active Inference agent was coded in the Julia programming language on a Raspberry Pi4-based Linux environment. {{< figure src="/img/projects/park-by-ai/robot.jpg" title="The robot that was used in the current study." width="400px" >}} -Figure 2 shows the information processing architecture of the active inference agent and its environmental interactions. The environment consists of a robot and an observer. The robot needs to park at a location that is initially unknown to the robot's steering agent (called "physics model" in the figure), but the observer _does_ know where to park. At any time, the observer can express binary (thumbs up or down) performance appraisals to the agent on the basis of her assessment on how well the robot is executing its parking task. The agent's task is to _learn_ the target parking location from these appraisals and steer the robot to this location. +Figure 2 shows the information processing architecture of the active inference agent and its environmental interactions. The environment consists of a robot and an observer. The robot needs to park at a location that is initially unknown to the robot's steering agent (called "physics model" in the figure), but the observer _does_ know where to park. At any time, the observer can express binary (thumbs up or down) performance appraisals to the agent on the basis of her assessment on how well the robot is executing its parking task. The agent's task is to _learn_ the target parking location from these appraisals and steer the robot to this location. {{< figure src="/img/projects/park-by-ai/robot-architecture.png" title="Information processing architecture of agent and its environmental interactions." width="600px" >}} -The agent is constructed as an Active Inference agent. The central tenet of Active Inference is that every task, including perception, learning and decision making, is framed as a Bayesian inference task in a (biased) generative model of the agents sensory inputs, which comprise (noisy) measurements of the robot's position and the observer's appraisals. The agent's generative model consists of two interacting sub-models: one model for predicting the robot's position and the other model for the observer appraisals. The agent is capable of simulating the physical model forward in time so as to predict where the robot will be located as a function of its steering signals. Initially, the physics model has no expectations about where to park. However, the agent's observer model infers desired future locations from the appraisals and relays this information to the physics model. Thus, as time progresses the physics model acquires more specific information about desired future positions. Through Bayesian inference (technically: variational inference), the agent is now capable to infer the steering signals that are needed to fulfill its expectations about (desired) future locations. Crucially, all information processing in this agent is framed as a variational inference task, which is automatable with modern probabilistic programming tools. +The agent is constructed as an Active Inference agent. The central tenet of Active Inference is that every task, including perception, learning and decision making, is framed as a Bayesian inference task in a (biased) generative model of the agents sensory inputs, which comprise (noisy) measurements of the robot's position and the observer's appraisals. The agent's generative model consists of two interacting sub-models: one model for predicting the robot's position and the other model for the observer appraisals. The agent is capable of simulating the physical model forward in time so as to predict where the robot will be located as a function of its steering signals. Initially, the physics model has no expectations about where to park. However, the agent's observer model infers desired future locations from the appraisals and relays this information to the physics model. Thus, as time progresses the physics model acquires more specific information about desired future positions. Through Bayesian inference (technically: variational inference), the agent is now capable to infer the steering signals that are needed to fulfill its expectations about (desired) future locations. Crucially, all information processing in this agent is framed as a variational inference task, which is automatable with modern probabilistic programming tools. -Various generative models were created, iteratively refined and verified in simulations and then ported to the robot. Inference algorithms were automatically generated using probabilistic programming toolboxes such as ForneyLab [[1](#references)] and Turing [[4](#references)]. +Various generative models were created, iteratively refined and verified in simulations and then ported to the robot. Inference algorithms were automatically generated using probabilistic programming toolboxes such as ForneyLab [[1](#references)] and Turing [[4](#references)]. ## Results @@ -55,7 +55,7 @@ Figure 4 shows a typical evolution of the agent's belief about the target locati Active inference agents also seem to be robust when they are subjected to environmental perturbations. The video below demonstrates how the active inference agent immediately corrects a severe manual interruption and continues its path towards the target location. -These experiments provide support for the notion that active inference is a viable method for constructing synthetic agents that are capable of learning new goals in a dynamic world. More details about this project are available in a MSc thesis [[2](#references)]. +These experiments provide support for the notion that active inference is a viable method for constructing synthetic agents that are capable of learning new goals in a dynamic world. More details about this project are available in a MSc thesis [[2](#references)]. {{}} @@ -63,7 +63,6 @@ These experiments provide support for the notion that active inference is a viab 1. Marco Cox, Thijs van de Laar, and Bert de Vries. A factor graph approach to automated design of Bayesian signal processing algorithms. International Journal of Approximate Reasoning, 104:185–204, January 2019. 2. Burak Ergül. [A Real-World Implementation of Active Inference.](/pdf/msc/Ergul-2020-MSc-thesis-A-Real-World-Implementation-of-Active-Inference.pdf) Master’s thesis, Eindhoven University of Technology, April 2020. -3. Karl Friston. The free-energy principle: a unified brain theory? Nature Reviews Neuroscience, 11(2):127–138, 2010. +3. Karl Friston. The free-energy principle: a unified brain theory? Nature Reviews Neuroscience, 11(2):127–138, 2010. 4. Hong Ge, Kai Xu, and Zoubin Ghahramani. Turing: a language for flexible probabilistic inference. InInternational Conference on Artificial Intelligence and Statistics, AISTATS 2018, 9-11 April 2018, Playa Blanca, Lanzarote,Canary Islands, Spain, pages 1682–1690, 2018. 5. Parallax Inc. [Robot shield with arduino.](https://www.parallax.com/product/32335), 2020. Accessed: 2020-04-08. - diff --git a/content/project/simulating-football-players-decision-making-process.md b/content/project/simulating-football-players-decision-making-process.md index 3060ff1..65b5841 100644 --- a/content/project/simulating-football-players-decision-making-process.md +++ b/content/project/simulating-football-players-decision-making-process.md @@ -27,7 +27,7 @@ sort_position = 1 ## Problem Statement Football has been the world's most popular sport for decades, and over time the game has evolved significantly. In order to search for an edge, Football clubs have started to adopt promising data analytics techniques. Consequently, the evolution of the game is pushed even further [[1](#references)]. -The goal of this project was to build an assistive tool that could help coaches to simulate the actions of an “optimally behaving” defending team in response to attackers. We defined defensive objectives and expressed those in a cost function for the defending team. +The goal of this project was to build an assistive tool that could help coaches to simulate the actions of an “optimally behaving” defending team in response to attackers. We defined defensive objectives and expressed those in a cost function for the defending team. During this work we made the following contributions: 1. We formulate two defense objectives from spatiotemporal football models based on the literature. @@ -63,6 +63,3 @@ To illustrate the working of our model, we simulated a line-breaking pass situat 1. D. Memmert and R. Rein, “Match analysis, Big Data and tactics: current trends in elite soccer,” Dtsch. Z. Für Sportmed., vol. 2018, no. 03, pp. 65–72, Mar. 2018 2. J. Fernandez, F. C. Barcelona, and L. Bornn, “Wide Open Spaces: A statistical technique for measuring space creation in professional soccer,” p. 19, 2018. 3. W. Spearman, A. Basye, G. Dick, R. Hotovy, and P. Pop, “Physics-Based Modeling of Pass Probabilities in Soccer,” p. 15, 2017. - - - diff --git a/content/publication/Efficient-model-evidence-computation-ScaleFactor.md b/content/publication/Efficient-model-evidence-computation-ScaleFactor.md index b171a52..0bca92d 100644 --- a/content/publication/Efficient-model-evidence-computation-ScaleFactor.md +++ b/content/publication/Efficient-model-evidence-computation-ScaleFactor.md @@ -15,7 +15,7 @@ url_dataset = "" url_project = "" url_slides = "" url_video = "" -url_custom = [ +url_custom = [ {name="IEEE", url = "https://ieeexplore.ieee.org/abstract/document/9919250"} ] date = "2022-11-04" diff --git a/content/publication/a-probabilistic-approach-to-situated-acoustic-road-event-detection.md b/content/publication/a-probabilistic-approach-to-situated-acoustic-road-event-detection.md index 75d6e0e..e433b0c 100644 --- a/content/publication/a-probabilistic-approach-to-situated-acoustic-road-event-detection.md +++ b/content/publication/a-probabilistic-approach-to-situated-acoustic-road-event-detection.md @@ -19,4 +19,3 @@ url_video = "" [[authors]] id = "mark" +++ - diff --git a/content/publication/asc-from-few-examples.md b/content/publication/asc-from-few-examples.md index 59d62ca..0cb9aa2 100644 --- a/content/publication/asc-from-few-examples.md +++ b/content/publication/asc-from-few-examples.md @@ -27,4 +27,4 @@ url_custom = [{name="IEEE", url = "https://ieeexplore.ieee.org/abstract/document [[authors]] name = "Bert de Vries" id = "bert" -+++ \ No newline at end of file ++++ diff --git a/content/publication/automated_design.md b/content/publication/automated_design.md index 771029c..1f73fda 100644 --- a/content/publication/automated_design.md +++ b/content/publication/automated_design.md @@ -1,13 +1,13 @@ +++ -abstract = '''This thesis describes a framework for automated design and personalization of Bayesian signal processing algorithms, with a focus on hearing aids. A pervasive problem with the hearing aid personalization (fitting) process is that acoustic issues often reveal themselves in-situ, far away from the controlled environment of the audiologists office. Instead of returning to the audiologist, it would be greatly beneficial to a patients hearing experience if she could adjust the hearing aid setting under situated conditions with minimal effort. Patient appraisals could for example be communicated to a synthetic algorithm designer by the press of a button on her smart-phone, or a gesture recognized by her smartwatch. The synthetic designer should then respond with an improved hearing aid signal processing algorithm. +abstract = '''This thesis describes a framework for automated design and personalization of Bayesian signal processing algorithms, with a focus on hearing aids. A pervasive problem with the hearing aid personalization (fitting) process is that acoustic issues often reveal themselves in-situ, far away from the controlled environment of the audiologists office. Instead of returning to the audiologist, it would be greatly beneficial to a patients hearing experience if she could adjust the hearing aid setting under situated conditions with minimal effort. Patient appraisals could for example be communicated to a synthetic algorithm designer by the press of a button on her smart-phone, or a gesture recognized by her smartwatch. The synthetic designer should then respond with an improved hearing aid signal processing algorithm. -Towards this goal, the current thesis advocates a problem-based iterative approach to algorithm design, rigorously based on probabilistic reasoning. Conventional hearing aid fitting adheres to prescribed fitting rules that are developed for an average user in a controlled environment. In contrast, situated design demands techniques that enable personalization to a specific user in an unknown and volatile environment. In this setting, there is much uncertainty about the user and her encompassing soundscape. Probability theory offers a rigorous language for reasoning with this uncertainty. +Towards this goal, the current thesis advocates a problem-based iterative approach to algorithm design, rigorously based on probabilistic reasoning. Conventional hearing aid fitting adheres to prescribed fitting rules that are developed for an average user in a controlled environment. In contrast, situated design demands techniques that enable personalization to a specific user in an unknown and volatile environment. In this setting, there is much uncertainty about the user and her encompassing soundscape. Probability theory offers a rigorous language for reasoning with this uncertainty. -In this thesis the algorithm design process commences with an explicit problem statement, in the form of a probabilistic generative model. In the context of hearing aid algorithm design, this generative model includes a statement about personal hearing loss. Additional information, about e.g. user preferences or the encompassing soundscape, can be incorporated by further constraints on the model. Crucially, algorithms for signal processing, parameter estimation and relative model comparison can be automatically inferred by Bayesian reasoning on the generative model. +In this thesis the algorithm design process commences with an explicit problem statement, in the form of a probabilistic generative model. In the context of hearing aid algorithm design, this generative model includes a statement about personal hearing loss. Additional information, about e.g. user preferences or the encompassing soundscape, can be incorporated by further constraints on the model. Crucially, algorithms for signal processing, parameter estimation and relative model comparison can be automatically inferred by Bayesian reasoning on the generative model. -Automatic algorithm derivation is implemented with the use of probabilistic programming techniques. The ForneyLab toolbox presented in this thesis allows for a convenient generative model specification as a Forney-style factor graph. This modular, scalable and flexible graphical formulation enables automatic derivation of complex (approximate) Bayesian message passing algorithms. Resulting algorithms are returned as Julia code, which can be customized and (in principle) readily executed on low-power devices such as hearing aids. +Automatic algorithm derivation is implemented with the use of probabilistic programming techniques. The ForneyLab toolbox presented in this thesis allows for a convenient generative model specification as a Forney-style factor graph. This modular, scalable and flexible graphical formulation enables automatic derivation of complex (approximate) Bayesian message passing algorithms. Resulting algorithms are returned as Julia code, which can be customized and (in principle) readily executed on low-power devices such as hearing aids. -Towards the goal of personalized algorithm design, the generative model can be extended with prior beliefs about future desired states. These prior beliefs will lead to goal-directed behavior by a process known as active inference. More specifically, if future outcomes are constrained to predict a satisfied user, then algorithms inferred from the constrained generative model will attempt to resolve these expectations. This thesis presents an operational framework for governing these active inference processes, which aid the automation of situated design processes. +Towards the goal of personalized algorithm design, the generative model can be extended with prior beliefs about future desired states. These prior beliefs will lead to goal-directed behavior by a process known as active inference. More specifically, if future outcomes are constrained to predict a satisfied user, then algorithms inferred from the constrained generative model will attempt to resolve these expectations. This thesis presents an operational framework for governing these active inference processes, which aid the automation of situated design processes. Taken together, this thesis describes the foundational principles for automated design of adaptive, personalized signal processing algorithms. The presented ForneyLab probabilistic programming software package offers the tools to implement the described algorithm design principles in practice.''' abstract_short = "" @@ -29,4 +29,4 @@ url_video = "" [[authors]] name = "Thijs van de Laar" id = "thijs" -+++ \ No newline at end of file ++++ diff --git a/content/publication/bayesian-pure-tone-audiometry-through-active-learning-under-informed-priors.md b/content/publication/bayesian-pure-tone-audiometry-through-active-learning-under-informed-priors.md index ffd84be..b7d86f6 100644 --- a/content/publication/bayesian-pure-tone-audiometry-through-active-learning-under-informed-priors.md +++ b/content/publication/bayesian-pure-tone-audiometry-through-active-learning-under-informed-priors.md @@ -24,4 +24,4 @@ url_video = "" id = "bert" +++ -View the [online version](https://www.frontiersin.org/articles/10.3389/fdgth.2021.723348/full) of this article. \ No newline at end of file +View the [online version](https://www.frontiersin.org/articles/10.3389/fdgth.2021.723348/full) of this article. diff --git a/content/publication/chance-constrained-active-inference.md b/content/publication/chance-constrained-active-inference.md index 7d2410e..9c07811 100644 --- a/content/publication/chance-constrained-active-inference.md +++ b/content/publication/chance-constrained-active-inference.md @@ -27,4 +27,4 @@ url_video = "" name = "Ayça Özçelikkale" [[authors]] name = "Henk Wymeersch" -+++ \ No newline at end of file ++++ diff --git a/content/publication/epistemic-value-graphical-models.md b/content/publication/epistemic-value-graphical-models.md index c118cb1..60c3ca3 100644 --- a/content/publication/epistemic-value-graphical-models.md +++ b/content/publication/epistemic-value-graphical-models.md @@ -30,4 +30,4 @@ url_pdf = "/pdf/frontiers-RAI/frobt-09-794464.pdf" [[authors]] name = "Bert de Vries" id = "bert" -+++ \ No newline at end of file ++++ diff --git a/content/publication/forneylab-biologically-plausible-fem.md b/content/publication/forneylab-biologically-plausible-fem.md index 7fe98f9..c5b4274 100644 --- a/content/publication/forneylab-biologically-plausible-fem.md +++ b/content/publication/forneylab-biologically-plausible-fem.md @@ -31,4 +31,4 @@ url_video = "" [[authors]] name = "Bert de Vries" id = "bert" -+++ \ No newline at end of file ++++ diff --git a/content/publication/forneylab-fast-and-flexible.md b/content/publication/forneylab-fast-and-flexible.md index 3847b6d..43ef75c 100644 --- a/content/publication/forneylab-fast-and-flexible.md +++ b/content/publication/forneylab-fast-and-flexible.md @@ -26,4 +26,4 @@ url_video = "" [[authors]] name = "Bert de Vries" id = "bert" -+++ \ No newline at end of file ++++ diff --git a/content/publication/gaussian-process-based-amorization-of-vmp-ur.md b/content/publication/gaussian-process-based-amorization-of-vmp-ur.md index 893e464..18074b1 100644 --- a/content/publication/gaussian-process-based-amorization-of-vmp-ur.md +++ b/content/publication/gaussian-process-based-amorization-of-vmp-ur.md @@ -15,7 +15,7 @@ url_dataset = "" url_project = "" url_slides = "" url_video = "" -url_custom = [ +url_custom = [ {name="IEEE", url = "https://ieeexplore.ieee.org/abstract/document/9909688"}, {name="EURASIP", url = "https://eurasip.org/Proceedings/Eusipco/Eusipco2022/pdfs/0001517.pdf"} ] diff --git a/content/publication/mp-algos-dynamical-systems.md b/content/publication/mp-algos-dynamical-systems.md index 7038f58..2b3f100 100644 --- a/content/publication/mp-algos-dynamical-systems.md +++ b/content/publication/mp-algos-dynamical-systems.md @@ -1,16 +1,16 @@ +++ -abstract = '''Building models to understand patterns is one of the fundamental pursuits of science. Among such models, dynamical models are ubiquitous in most scientific fields as these models describe how processes evolve. These models can be found in finance, navigation, control engineering, audio signal processing, and telecommunications. Applications range from tracking the position of an aircraft to estimating the variance of returns on assets. Inference corresponds to computing posterior distributions over the variables of a given model. +abstract = '''Building models to understand patterns is one of the fundamental pursuits of science. Among such models, dynamical models are ubiquitous in most scientific fields as these models describe how processes evolve. These models can be found in finance, navigation, control engineering, audio signal processing, and telecommunications. Applications range from tracking the position of an aircraft to estimating the variance of returns on assets. Inference corresponds to computing posterior distributions over the variables of a given model. This dissertation describes a theoretical framework for deriving customized message passing-based inference algorithms in factor graphs and illustrates the framework’s application to hierarchical dynamical models. Factor graphs are visual representations of the dependency structures among the variables of a model. Inference tasks on a given model can be realized using message passing algorithms on the corresponding factor graph, where propagated messages are computed by integration (summation). Often, dynamical models of natural processes are constructed hierarchically. Because the hierarchies in the models may grow the complexity of dependence structures, exact inference by message passing in these models becomes infeasible and computationally impossible in a real-time setting. To employ hierarchical dynamical models in applications that require real-time processing, inference by message passing needs to be approximated. This dissertation proposes a constraint manipulation strategy to derive message passing algorithms on Forney-style factor graphs that pave the way for an efficient implementation of automated approximate inference. By changing the constraints on the local sub-graphs, we derive various local message update rules as the stationary solutions of a constrained Bethe free energy from first principles. By combining these local updates, one can perform hybrid message passing. Constraint manipulation is a modular way of generating message-passing algorithms by combining local updates such that factorized computations of local updates allow efficient implementation. This thesis then demonstrates how message passing by constraint manipulation applies to -hierarchical dynamical systems. The focus is on the hierarchical Gaussian filter, a time-series model for volatile processes where non-linear transforms couple the states in this process. A composite factor node (named GCV), representing the state transition distribution of an HGF, is constructed and subsequently can be used as a plug-in module for any factor graph. Various message update rules for the GCV node under multiple constraints are derived. Combining derived update rules, it is possible to implement automated hybrid message passing for the variants of the HGF-like models in software packages ForneyLab.jl and ReactiveMP.jl. +hierarchical dynamical systems. The focus is on the hierarchical Gaussian filter, a time-series model for volatile processes where non-linear transforms couple the states in this process. A composite factor node (named GCV), representing the state transition distribution of an HGF, is constructed and subsequently can be used as a plug-in module for any factor graph. Various message update rules for the GCV node under multiple constraints are derived. Combining derived update rules, it is possible to implement automated hybrid message passing for the variants of the HGF-like models in software packages ForneyLab.jl and ReactiveMP.jl. Natural processes are often non-stationary. Therefore, the realizations of natural processes have statistical properties that change with time. A source of non-stationarity is due to regime changes in the parameter values. A parameterized transition distribution may govern changes in the statistical properties. If the parameters of this transition distribution are subject to regime switches, then the statistical properties of the transition distribution will depend on the regimes. To account for context switches, this dissertation provides a switching extension to the HGF model with a hidden Markov model that governs a selection mechanism for the parameters of the ordinary HGF model. A composite factor node (named GCSV) is -constructed as a successor of GCV, and closed-form message update rules are derived. The derived message update rules allow automated real-time message passing in graphs containing state transitions with switching volatile dynamics. +constructed as a successor of GCV, and closed-form message update rules are derived. The derived message update rules allow automated real-time message passing in graphs containing state transitions with switching volatile dynamics. Moreover, this dissertation illustrates how the graphical formalism of factor graphs allows us to build complex models from primitive node structures. To that end, this dissertation focuses on auto-regressive(AR) processes that are ubiquitous for time-series modeling. AR processes are often constructed under the assumption that the precision of the innovation noise and AR coefficients are constant to ensure stationarity. The dissertation shows how the GCV node could be used as a plug-in module within the graphs of auto-regressive models to extend the auto-regressive models such that the deriving noise processes are time-varying. Message passing in the corresponding model leads to online state and parameter estimation in auto-regressive models with time-varying process noise. @@ -37,4 +37,4 @@ url_video = "" [[authors]] id = "ismail" -+++ \ No newline at end of file ++++ diff --git a/content/publication/mp-based-har.md b/content/publication/mp-based-har.md index 93742af..2c724cd 100644 --- a/content/publication/mp-based-har.md +++ b/content/publication/mp-based-har.md @@ -1,7 +1,7 @@ +++ -abstract = '''This dissertation describes a research effort toward automating personalized design of hearing aid algorithms through in-the-field communication between a user and a portable intelligent agent. The traditional design cycle of hearing aid is inefficient as it requires many human professionals in the design loop who have to elicit and design for a hearing impaired person's unique and context-dependent preferences. In contrast, a wearable synthetic intelligent agent could possibly improve the quality of a hearing aid by on-the-spot suggestions for new hearing aid settings, rather than waiting for offline human expert intervention. To create such an agent, we take inspiration from a theoretical neuroscience framework called the Free Energy Principle, which explains how living brains effectively control their environment by online Bayesian learning of a model of their environment. +abstract = '''This dissertation describes a research effort toward automating personalized design of hearing aid algorithms through in-the-field communication between a user and a portable intelligent agent. The traditional design cycle of hearing aid is inefficient as it requires many human professionals in the design loop who have to elicit and design for a hearing impaired person's unique and context-dependent preferences. In contrast, a wearable synthetic intelligent agent could possibly improve the quality of a hearing aid by on-the-spot suggestions for new hearing aid settings, rather than waiting for offline human expert intervention. To create such an agent, we take inspiration from a theoretical neuroscience framework called the Free Energy Principle, which explains how living brains effectively control their environment by online Bayesian learning of a model of their environment. -According to this hypothesis, an agent (such as a brain) holds a generative probabilistic model for its sensory input signals. Translated to the context of a synthetic agent and an acoustic environment with a hearing aid (HA) and a HA patient, the agent's generative model should comprise a model for both environmental acoustic signals and user appraisals for hearing aid behavior. These models ought to be learned under in-situ conditions through Bayesian inference, which offers a rigorous procedure for parameter estimation in probabilistic models. +According to this hypothesis, an agent (such as a brain) holds a generative probabilistic model for its sensory input signals. Translated to the context of a synthetic agent and an acoustic environment with a hearing aid (HA) and a HA patient, the agent's generative model should comprise a model for both environmental acoustic signals and user appraisals for hearing aid behavior. These models ought to be learned under in-situ conditions through Bayesian inference, which offers a rigorous procedure for parameter estimation in probabilistic models. Following the premise of the Free Energy Principle, the essence of our approach to automated HA design is that all engineering tasks can be formulated as a Bayesian inference on the generative probabilistic model. In particular, this dissertation focuses on a specific family of models for environmental acoustical signals, namely Hierarchical Autoregressive Models. In principle, the flexibility of these models supports describing complex non-stationary acoustic environments. Unfortunately, Bayesian parameter estimation in these models is not trivial, and inference solutions do not exist in closed-form. Therefore, this work develops methods to automate Bayesian inference for both state and parameter updating in hierarchical autoregressive models. @@ -28,4 +28,4 @@ url_video = "" [[authors]] name = "Albert Podusenko" id = "albert" -+++ \ No newline at end of file ++++ diff --git a/content/publication/realising_synthetic_active_inference_agents,_part_i:_epistemic_objectives_and_graphical_specification_language.md b/content/publication/realising_synthetic_active_inference_agents_part_i_epistemic_objectives_and_graphical_specification_language.md similarity index 100% rename from content/publication/realising_synthetic_active_inference_agents,_part_i:_epistemic_objectives_and_graphical_specification_language.md rename to content/publication/realising_synthetic_active_inference_agents_part_i_epistemic_objectives_and_graphical_specification_language.md diff --git a/content/publication/realising_synthetic_active_inference_agents,_part_ii:_variational_message_updates.md b/content/publication/realising_synthetic_active_inference_agents_part_ii_variational_message_updates.md similarity index 100% rename from content/publication/realising_synthetic_active_inference_agents,_part_ii:_variational_message_updates.md rename to content/publication/realising_synthetic_active_inference_agents_part_ii_variational_message_updates.md diff --git a/content/publication/robust-expectation-propagation.md b/content/publication/robust-expectation-propagation.md index f392720..d46d85d 100644 --- a/content/publication/robust-expectation-propagation.md +++ b/content/publication/robust-expectation-propagation.md @@ -23,4 +23,4 @@ url_custom = [{name="IEEE", url = "https://ieeexplore.ieee.org/document/8553490" [[authors]] name = "Bert de Vries" id = "bert" -+++ \ No newline at end of file ++++ diff --git a/content/publication/simulating-active-inference.md b/content/publication/simulating-active-inference.md index b22e755..439980e 100644 --- a/content/publication/simulating-active-inference.md +++ b/content/publication/simulating-active-inference.md @@ -22,4 +22,4 @@ url_video = "" [[authors]] name = "Bert de Vries" id = "bert" -+++ \ No newline at end of file ++++ diff --git a/content/publication/the-graphical-brain.md b/content/publication/the-graphical-brain.md index 6254a71..25925c5 100644 --- a/content/publication/the-graphical-brain.md +++ b/content/publication/the-graphical-brain.md @@ -19,7 +19,7 @@ url_video = "" [[authors]] name = "Karl J Friston" [[authors]] - name = "Thomas Parr" + name = "Thomas Parr" [[authors]] name = "Bert de Vries" id = "bert" diff --git a/content/publication/universal-probprog-with-MP-on-FGs.md b/content/publication/universal-probprog-with-MP-on-FGs.md index 5cc2adf..56a11c8 100644 --- a/content/publication/universal-probprog-with-MP-on-FGs.md +++ b/content/publication/universal-probprog-with-MP-on-FGs.md @@ -44,4 +44,4 @@ url_video = "" [[authors]] name = "Semih Akbayrak" id = "semih" -+++ \ No newline at end of file ++++ diff --git a/content/publication/worked-example-fokker.md b/content/publication/worked-example-fokker.md index 62013e9..ddd7650 100644 --- a/content/publication/worked-example-fokker.md +++ b/content/publication/worked-example-fokker.md @@ -1,5 +1,5 @@ +++ -abstract = "The Free Energy Principle (FEP) and its corollary active inference describe a complex theoretical framework with a substantial statistical mechanics foundation that is often expressed in terms of the Fokker-Planck equation. Easy-to-follow examples of this formalism are scarce, leaving a high barrier of entry to the field. In this paper we provide a worked example of an active inference agent as a hierarchical Gaussian generative model. We proceed to write its equations of motion explicitly as a Fokker-Planck equation, providing a clear mapping between theoretical accounts of FEP and practical implementation." +abstract = "The Free Energy Principle (FEP) and its corollary active inference describe a complex theoretical framework with a substantial statistical mechanics foundation that is often expressed in terms of the Fokker-Planck equation. Easy-to-follow examples of this formalism are scarce, leaving a high barrier of entry to the field. In this paper we provide a worked example of an active inference agent as a hierarchical Gaussian generative model. We proceed to write its equations of motion explicitly as a Fokker-Planck equation, providing a clear mapping between theoretical accounts of FEP and practical implementation." abstract_short = "" date = "2020-09-14T10:40:00+02:00" image = "" diff --git a/content/teaching/archive/aip-5ssb0.md b/content/teaching/archive/aip-5ssb0.md index 1ee5973..33245da 100644 --- a/content/teaching/archive/aip-5ssb0.md +++ b/content/teaching/archive/aip-5ssb0.md @@ -36,7 +36,7 @@ and **starts on 4-Feb-2019**. ## News -- **26-Mar-2019** We added an extra old exam (from April 2017, [with solutions](https://github.com/bertdv/AIP-5SSB0/raw/master/lessons/exercises/170410-5SSB0-exam-with-solutions.pdf)) to aid your exam preparation. +- **26-Mar-2019** We added an extra old exam (from April 2017, [with solutions](https://github.com/bertdv/AIP-5SSB0/raw/master/lessons/exercises/170410-5SSB0-exam-with-solutions.pdf)) to aid your exam preparation. - **12-Mar-2019** The [pdf handout for part-2](https://github.com/bertdv/AIP-5SSB0/raw/master/lessons/Tjalling/AIP-part2-handout.pdf) has been updated. - **5-Feb-2019** The [pdf bundle for part-1](https://github.com/bertdv/AIP-5SSB0/raw/master/output/AIP-5SSB0.pdf) has been updated. @@ -85,8 +85,8 @@ Markov models and various latent component analysis models. Furthermore, we deri - [8 - Discriminative Classification](http://nbviewer.ipython.org/github/bertdv/AIP-5SSB0/blob/master/lessons/notebooks/08_Discriminative-Classification.ipynb) - [9 - Clustering with Gaussian Mixture Models](http://nbviewer.ipython.org/github/bertdv/AIP-5SSB0/blob/master/lessons/notebooks/09_Clustering-with-Gaussian-Mixture-Models.ipynb) - [10- The EM Algorithm](http://nbviewer.ipython.org/github/bertdv/AIP-5SSB0/blob/master/lessons/notebooks/10_The-General-EM-Algorithm.ipynb) -- [11- Continuous Latent Variable Models - PCA and FA](http://nbviewer.ipython.org/github/bertdv/AIP-5SSB0/blob/master/lessons/notebooks/11_Continuous-Latent-Variable-Models-PCA-and-FA.ipynb) -- [12- Dynamic Latent Variable Models](http://nbviewer.ipython.org/github/bertdv/AIP-5SSB0/blob/master/lessons/notebooks/12_Dynamic-Latent-Variable-Models.ipynb) +- [11- Continuous Latent Variable Models - PCA and FA](http://nbviewer.ipython.org/github/bertdv/AIP-5SSB0/blob/master/lessons/notebooks/11_Continuous-Latent-Variable-Models-PCA-and-FA.ipynb) +- [12- Dynamic Latent Variable Models](http://nbviewer.ipython.org/github/bertdv/AIP-5SSB0/blob/master/lessons/notebooks/12_Dynamic-Latent-Variable-Models.ipynb) - [13- Factor Graphs and Message Passing Algorithms](http://nbviewer.ipython.org/github/bertdv/AIP-5SSB0/blob/master/lessons/notebooks/13_Factor-Graphs-and-Message-Passing-Algorithms.ipynb) diff --git a/content/teaching/archive/bmlip-2019.md b/content/teaching/archive/bmlip-2019.md index 47d250a..6329bf7 100644 --- a/content/teaching/archive/bmlip-2019.md +++ b/content/teaching/archive/bmlip-2019.md @@ -43,9 +43,9 @@ After the first question, the rest of the exam will be more focused at selected ## Instructors -- Instructors: [prof.dr.ir. Bert de Vries](http://bertdv.nl) (responsible instructor) and [dr. Wouter Kouw](https://biaslab.github.io/member/wouter/). +- Instructors: [prof.dr.ir. Bert de Vries](http://bertdv.nl) (responsible instructor) and [dr. Wouter Kouw](https://biaslab.github.io/member/wouter/). -- Teaching assistants: [Ismail Senoz, MSc](https://biaslab.github.io/member/ismail/), and [Magnus Koudahl, MSc](https://biaslab.github.io/member/magnus/). +- Teaching assistants: [Ismail Senoz, MSc](https://biaslab.github.io/member/ismail/), and [Magnus Koudahl, MSc](https://biaslab.github.io/member/magnus/). @@ -95,16 +95,16 @@ Learning](https://www.microsoft.com/en-us/research/people/cmbishop/prml-book/). #### Minicourse Probabilistic Programming -- 13- Linear Regression & Classification - - [with Monte Carlo Sampling](http://nbviewer.ipython.org/github/bertdv/BMLIP/blob/master/lessons/notebooks/probprog/Probabilistic-Programming-1-sampling.ipynb) +- 13- Linear Regression & Classification + - [with Monte Carlo Sampling](http://nbviewer.ipython.org/github/bertdv/BMLIP/blob/master/lessons/notebooks/probprog/Probabilistic-Programming-1-sampling.ipynb) - [with Variational Inference](http://nbviewer.ipython.org/github/bertdv/BMLIP/blob/master/lessons/notebooks/probprog/Probabilistic-Programming-1-variational.ipynb) -- 14- Gaussian Mixture Model +- 14- Gaussian Mixture Model - [with Monte Carlo Sampling](http://nbviewer.ipython.org/github/bertdv/BMLIP/blob/master/lessons/notebooks/probprog/Probabilistic-Programming-2-sampling.ipynb) - - [with Variational Inference](http://nbviewer.ipython.org/github/bertdv/BMLIP/blob/master/lessons/notebooks/probprog/Probabilistic-Programming-2-variational.ipynb) -- 15- Hidden Markov Model - - [with Monte Carlo Sampling](http://nbviewer.ipython.org/github/bertdv/BMLIP/blob/master/lessons/notebooks/probprog/Probabilistic-Programming-3-sampling.ipynb) - - [with Variational Inference](http://nbviewer.ipython.org/github/bertdv/BMLIP/blob/master/lessons/notebooks/probprog/Probabilistic-Programming-3-variational.ipynb) -- 16- Kalman Filtering + - [with Variational Inference](http://nbviewer.ipython.org/github/bertdv/BMLIP/blob/master/lessons/notebooks/probprog/Probabilistic-Programming-2-variational.ipynb) +- 15- Hidden Markov Model + - [with Monte Carlo Sampling](http://nbviewer.ipython.org/github/bertdv/BMLIP/blob/master/lessons/notebooks/probprog/Probabilistic-Programming-3-sampling.ipynb) + - [with Variational Inference](http://nbviewer.ipython.org/github/bertdv/BMLIP/blob/master/lessons/notebooks/probprog/Probabilistic-Programming-3-variational.ipynb) +- 16- Kalman Filtering - [with Monte Carlo sampling](http://nbviewer.ipython.org/github/bertdv/BMLIP/blob/master/lessons/notebooks/probprog/Probabilistic-Programming-4-sampling.ipynb) - [with Variational Inference](http://nbviewer.ipython.org/github/bertdv/BMLIP/blob/master/lessons/notebooks/probprog/Probabilistic-Programming-4-variational.ipynb) diff --git a/content/teaching/archive/bmlip-2020.md b/content/teaching/archive/bmlip-2020.md index 29c54cc..2c2e944 100644 --- a/content/teaching/archive/bmlip-2020.md +++ b/content/teaching/archive/bmlip-2020.md @@ -1,8 +1,8 @@ +++ date = "2018-08-23T14:45:00+02:00" external_link = "" -title = "Bayesian Machine Learning and Information Processing (5SSD0)" -subtitle = "academic year 2020/21" +title = "Bayesian Machine Learning and Information Processing (5SSD0)" +subtitle = "academic year 2020/21" participants_block_position = "down" type = "teaching" @@ -23,13 +23,13 @@ type = "teaching" +++ - -The 2020/21 course "Bayesian Machine Learning and Information Processing" will start in November 2020 (Q2). +The 2020/21 course "Bayesian Machine Learning and Information Processing" will start in November 2020 (Q2). @@ -44,14 +44,14 @@ This course provides an introduction to Bayesian machine learning and informatio News and Announcements -- As much as possible we use the [Piazza course site](https://piazza.com/class/kgp8llbdmx84s9) for new announcements. +- As much as possible we use the [Piazza course site](https://piazza.com/class/kgp8llbdmx84s9) for new announcements. ## Instructors - [Prof.dr.ir. Bert de Vries](http://bertdv.nl) (email: bert.de.vries@tue.nl) is the responsible instructor for this course and teaches all [lectures with label B](#lectures). -- [Dr. Wouter Kouw](https://biaslab.github.io/member/wouter/) (w.m.kouw@tue.nl) teaches all [practical sessions on probabilistic programming with label W](#lectures). -- [Ismail Senoz, MSc](https://biaslab.github.io/member/ismail/) (i.senoz@tue.nl), and [Magnus Koudahl, MSc](https://biaslab.github.io/member/magnus/) (m.t.koudahl@tue.nl) are teaching assistants. Mr. Koudahl presents the ["What is Life?"](#bonus-lecture) bonus lecture. +- [Dr. Wouter Kouw](https://biaslab.github.io/member/wouter/) (w.m.kouw@tue.nl) teaches all [practical sessions on probabilistic programming with label W](#lectures). +- [Ismail Senoz, MSc](https://biaslab.github.io/member/ismail/) (i.senoz@tue.nl), and [Magnus Koudahl, MSc](https://biaslab.github.io/member/magnus/) (m.t.koudahl@tue.nl) are teaching assistants. Mr. Koudahl presents the ["What is Life?"](#bonus-lecture) bonus lecture. ## Materials @@ -67,9 +67,9 @@ Bishop](http://research.microsoft.com/en-us/um/people/cmbishop/index.htm) (2006) Machine Learning](https://www.microsoft.com/en-us/research/people/cmbishop/prml-book/). You can also buy a [hardcopy, e.g. at bol.com](https://tinyurl.com/thj7euq). 2. [Ariel Caticha](https://www.albany.edu/physics/acaticha.shtml) (2012), [Entropic Inference and the Foundations of Physics](https://github.com/bertdv/BMLIP/blob/master/lessons/notebooks/files/Caticha-2012-Entropic-Inference-and-the-Foundations-of-Physics.pdf). -3. Bert de Vries et al. (2020), [PDF bundle of lecture notes for lessons B0 through B12 (Ed. Q3-2019/20)](https://github.com/bertdv/BMLIP/blob/master/lessons/notebooks/files/5SSD0-Mar2020-Lecture-notes.pdf?dl=0). +3. Bert de Vries et al. (2020), [PDF bundle of lecture notes for lessons B0 through B12 (Ed. Q3-2019/20)](https://github.com/bertdv/BMLIP/blob/master/lessons/notebooks/files/5SSD0-Mar2020-Lecture-notes.pdf?dl=0). - The lecture notes may change a bit during the course, e.g., to process comments by students. A final PDF version will be posted after the last lecture. -4. Wouter Kouw (2020), [Julia and Jupyter Install Guide](https://github.com/bertdv/BMLIP/blob/master/lessons/notebooks/files/WKouw-Mar2020-JuliaJupyterInstallGuide.pdf?dl=0). +4. Wouter Kouw (2020), [Julia and Jupyter Install Guide](https://github.com/bertdv/BMLIP/blob/master/lessons/notebooks/files/WKouw-Mar2020-JuliaJupyterInstallGuide.pdf?dl=0). - Use this guide if you need help to install [Julia](https://julialang.org) and [Jupyter](https://jupyter.org/), so that you can open and run the course notebooks on your own machine. - You can test your installation by running the notebook called "Probabilistic-Programming-0.ipynb", which can be downloaded from [github](https://github.com/bertdv/bmlip) (under `lessons/notebooks/probprog`). [Here](https://www.youtube.com/watch?v=BWGTudg3xlI) is a video with step-by-step instructions on opening course notebooks. @@ -228,18 +228,15 @@ Please feel free to consult the following matrix and Gaussian cheat sheets (by S ## Exam Guide -Each year there will be two exam opportunities. Check the official TUE course site for exam schedules. In the Q2-2020 course, your performance will be assessed by a WRITTEN EXAMINATION, which (very likely) will be offered both online (with proctoring software) and offline (on campus, if the situation allows it). +Each year there will be two exam opportunities. Check the official TUE course site for exam schedules. In the Q2-2020 course, your performance will be assessed by a WRITTEN EXAMINATION, which (very likely) will be offered both online (with proctoring software) and offline (on campus, if the situation allows it). **You cannot bring notes or books to the exam. All needed formulas are supplied at the exam sheet**. - - - - diff --git a/content/teaching/archive/bmlip-2021.md b/content/teaching/archive/bmlip-2021.md index cf51d8e..4d02994 100644 --- a/content/teaching/archive/bmlip-2021.md +++ b/content/teaching/archive/bmlip-2021.md @@ -1,8 +1,8 @@ +++ date = "2018-08-23T14:45:00+02:00" external_link = "" -title = "Bayesian Machine Learning and Information Processing (5SSD0)" -subtitle = "academic year 2021/22" +title = "Bayesian Machine Learning and Information Processing (5SSD0)" +subtitle = "academic year 2021/22" participants_block_position = "down" type = "teaching" @@ -35,7 +35,7 @@ Note: This site is currently under construction. ---> -The 2021/22 course "Bayesian Machine Learning and Information Processing" will start in November 2021 (Q2). +The 2021/22 course "Bayesian Machine Learning and Information Processing" will start in November 2021 (Q2). @@ -58,18 +58,18 @@ News and Announcements - 01-Dec-2021: Last year's Probabilistic Programming assignments have been made available as exercises. Solutions are given as well. -- 26-Nov-2021: As per the TU/e mandate, there will be no assignment given prior to the Christmas break. +- 26-Nov-2021: As per the TU/e mandate, there will be no assignment given prior to the Christmas break. -- 13-Nov-2021: This year's live classes will be online! +- 13-Nov-2021: This year's live classes will be online! -- As much as possible we use the [Piazza course site](https://piazza.com/class/ku2o9c3f71a4px) for new announcements. +- As much as possible we use the [Piazza course site](https://piazza.com/class/ku2o9c3f71a4px) for new announcements. ## Instructors - [Prof.dr.ir. Bert de Vries](http://bertdv.nl) (email: bert.de.vries@tue.nl) is the responsible instructor for this course and teaches all [lectures with label B](#lectures). -- [Dr. Wouter Kouw](https://biaslab.github.io/member/wouter/) (w.m.kouw@tue.nl) teaches all [practical sessions on probabilistic programming with label W](#lectures). -- [Magnus Koudahl, MSc](https://biaslab.github.io/member/magnus/) (m.t.koudahl@tue.nl) is the teaching assistant. Mr. Koudahl presents the ["What is Life?"](#bonus-lecture) bonus lecture. +- [Dr. Wouter Kouw](https://biaslab.github.io/member/wouter/) (w.m.kouw@tue.nl) teaches all [practical sessions on probabilistic programming with label W](#lectures). +- [Magnus Koudahl, MSc](https://biaslab.github.io/member/magnus/) (m.t.koudahl@tue.nl) is the teaching assistant. Mr. Koudahl presents the ["What is Life?"](#bonus-lecture) bonus lecture. ## Materials @@ -80,7 +80,7 @@ In principle, you can download all needed materials from the links below. Please consider downloading the following books/resources: -- Bert de Vries (2021), [PDF bundle of all lecture notes for lessons B0 through B12](https://github.com/bertdv/BMLIP/blob/a75b6f0f12fa393cfdd61526052ee42559f01f4e/output/BMLIP-5SSD0-lecture-notes.pdf). +- Bert de Vries (2021), [PDF bundle of all lecture notes for lessons B0 through B12](https://github.com/bertdv/BMLIP/blob/a75b6f0f12fa393cfdd61526052ee42559f01f4e/output/BMLIP-5SSD0-lecture-notes.pdf). - Wouter Kouw (2021), [PDF bundle of all probabilistic programming lecture notes for lessons W1 through W4](https://github.com/bertdv/BMLIP/blob/a75b6f0f12fa393cfdd61526052ee42559f01f4e/output/BMLIP-5SSD0-prob-prog.pdf). - [Christopher M. Bishop](http://research.microsoft.com/en-us/um/people/cmbishop/index.htm) (2006), [Pattern Recognition and @@ -90,8 +90,8 @@ Learning](https://www.microsoft.com/en-us/research/people/cmbishop/prml-book/). ### Software -- The [software installation guide](https://github.com/bertdv/BMLIP/blob/a75b6f0f12fa393cfdd61526052ee42559f01f4e/lessons/notebooks/files/BMLIP-installation-guide.pdf) contains step-by-step instructions to setup everything you need to run the course notebooks yourself. -- You can test your installation by running the notebook called [Probabilistic-Programming-0.ipynb](https://github.com/bertdv/BMLIP/blob/a75b6f0f12fa393cfdd61526052ee42559f01f4e/lessons/notebooks/probprog/Probabilistic-Programming-0.ipynb) which can be found in the ([github](https://github.com/bertdv/bmlip)) repo under `lessons/notebooks/probprog`. +- The [software installation guide](https://github.com/bertdv/BMLIP/blob/a75b6f0f12fa393cfdd61526052ee42559f01f4e/lessons/notebooks/files/BMLIP-installation-guide.pdf) contains step-by-step instructions to setup everything you need to run the course notebooks yourself. +- You can test your installation by running the notebook called [Probabilistic-Programming-0.ipynb](https://github.com/bertdv/BMLIP/blob/a75b6f0f12fa393cfdd61526052ee42559f01f4e/lessons/notebooks/probprog/Probabilistic-Programming-0.ipynb) which can be found in the ([github](https://github.com/bertdv/bmlip)) repo under `lessons/notebooks/probprog`. ### Lecture notes, videos and exercises @@ -141,7 +141,7 @@ You can access all lecture notes, videos and exercises online through the links W1: Probabilistic Programming 1 - Intro Bayesian ML W1.1, W1.2, W1.3 W1 - W1-ex
+ W1-ex
W1-sol W1 @@ -174,7 +174,7 @@ You can access all lecture notes, videos and exercises online through the links W2: ProbProg 2 - MP & Analytical Bayesian Solutions W2.1, W2.2, W2.3 W2 - W2-ex
+ W2-ex
W2-sol W2 @@ -211,7 +211,7 @@ You can access all lecture notes, videos and exercises online through the links W3: ProbProg 3 - Regression and Classification W3.1, W3.2 W3 - W3-ex
+ W3-ex
W3-sol W3 @@ -236,7 +236,7 @@ You can access all lecture notes, videos and exercises online through the links W4: ProbProg 4: Latent Variable and Dynamic Models W4.1, W4.2 W4 - W4-ex
+ W4-ex
W4-sol W4 @@ -302,26 +302,23 @@ Please feel free to consult the following matrix and Gaussian cheat sheets (by S - [Gaussian Identities](https://github.com/bertdv/BMLIP/raw/a75b6f0f12fa393cfdd61526052ee42559f01f4e/lessons/notebooks/files/Roweis-1999-gaussian-identities.pdf?dl=0) - [Matrix Identities](https://github.com/bertdv/BMLIP/raw/a75b6f0f12fa393cfdd61526052ee42559f01f4e/lessons/notebooks/files/Roweis-1999-matrix-identities.pdf?dl=0) -## Class logistics +## Class logistics Here's our recommendation on how to study for this class. Before each lecture: - first watch the video guide for that lecture (2nd column in above table) - then study the lecture notes (3rd column) - then (optionally), watch the live class recording from the previous (2020/21) edition (in 4th colum) - then try to make the exercises (5th column) for that class. Feel free to use this cheatsheet to make the exercises. - - If you have any remaining issues or questions, please pose your question in piazza. Your questions will be answered at the piazza site by fellow students and accorded (or corrected, amended) by the teaching staff. + - If you have any remaining issues or questions, please pose your question in piazza. Your questions will be answered at the piazza site by fellow students and accorded (or corrected, amended) by the teaching staff. -The live classes in this edition of 5SSD0 are (optional) flipped classroom-style Q&A sessions, where we discuss any issues in person. This year the live class sessions are not recorded so you will not be able to view them later. We do not intend to present new materials in the live classes, so in principle, the materials in the above table should suffice as preparation for the written exam. +The live classes in this edition of 5SSD0 are (optional) flipped classroom-style Q&A sessions, where we discuss any issues in person. This year the live class sessions are not recorded so you will not be able to view them later. We do not intend to present new materials in the live classes, so in principle, the materials in the above table should suffice as preparation for the written exam. ---> - - - - diff --git a/content/teaching/archive/bmlip-2022.md b/content/teaching/archive/bmlip-2022.md index ac10225..d6cbffb 100644 --- a/content/teaching/archive/bmlip-2022.md +++ b/content/teaching/archive/bmlip-2022.md @@ -29,7 +29,7 @@ Note: This site is currently under construction. ---> -The 2022/23 course "Bayesian Machine Learning and Information Processing" will start in November 2022 (Q2). +The 2022/23 course "Bayesian Machine Learning and Information Processing" will start in November 2022 (Q2). @@ -48,18 +48,18 @@ News and Announcements - The solution notebook for the Probabilistic Programming assignment can be downloaded [here](https://github.com/bertdv/BMLIP/blob/master/lessons/assignment/Probabilistic%20Programming%20-%20Assignment%20%5BSolution%5D.ipynb). -- [MSc projects overview at BIASlab for download](https://github.com/bertdv/BMLIP/blob/master/lessons/notebooks/files/MScprojects-2023.pdf) +- [MSc projects overview at BIASlab for download](https://github.com/bertdv/BMLIP/blob/master/lessons/notebooks/files/MScprojects-2023.pdf) -- Exam rules have been [posted at piazza](https://piazza.com/class/l9n5gnieu4k6tl/post/23) +- Exam rules have been [posted at piazza](https://piazza.com/class/l9n5gnieu4k6tl/post/23) -- As much as possible we use the [Piazza course site](https://piazza.com/class/l9n5gnieu4k6tl) for new announcements. +- As much as possible we use the [Piazza course site](https://piazza.com/class/l9n5gnieu4k6tl) for new announcements. ## Instructors - [Prof.dr.ir. Bert de Vries](http://bertdv.nl) (email: bert.de.vries@tue.nl) is the responsible instructor for this course and teaches all [lectures with label B](#lectures). -- [Dr. Wouter Kouw](https://biaslab.github.io/member/wouter/) (w.m.kouw@tue.nl) teaches all [practical sessions on probabilistic programming with label W](#lectures). -- [Magnus Koudahl](https://biaslab.github.io/member/magnus/), [Tim Nisslbeck](https://biaslab.github.io/member/tim), [Sepideh Adamiat](https://biaslab.github.io/member/sepideh) and [Wouter Nuijten](https://biaslab.github.io/member/woutern) are the teaching assistants. Mr. Koudahl presents the ["What is Life?"](#bonus-lecture) bonus lecture. +- [Dr. Wouter Kouw](https://biaslab.github.io/member/wouter/) (w.m.kouw@tue.nl) teaches all [practical sessions on probabilistic programming with label W](#lectures). +- [Magnus Koudahl](https://biaslab.github.io/member/magnus/), [Tim Nisslbeck](https://biaslab.github.io/member/tim), [Sepideh Adamiat](https://biaslab.github.io/member/sepideh) and [Wouter Nuijten](https://biaslab.github.io/member/woutern) are the teaching assistants. Mr. Koudahl presents the ["What is Life?"](#bonus-lecture) bonus lecture. ## Materials @@ -69,7 +69,7 @@ In principle, you can download all needed materials from the links below. Please consider downloading the following books/resources: -- Bert de Vries (2022), [PDF bundle of all lecture notes for lessons B0 through B12](https://github.com/bertdv/BMLIP/blob/master/output/BMLIP-5SSD0-lecture-notes.pdf). +- Bert de Vries (2022), [PDF bundle of all lecture notes for lessons B0 through B12](https://github.com/bertdv/BMLIP/blob/master/output/BMLIP-5SSD0-lecture-notes.pdf). - Wouter Kouw (2022), [PDF bundle of all probabilistic programming lecture notes for lessons W1 through W4](https://github.com/bertdv/BMLIP/blob/master/output/BMLIP-5SSD0-prob-prog.pdf). - [Christopher M. Bishop](http://research.microsoft.com/en-us/um/people/cmbishop/index.htm) (2006), [Pattern Recognition and @@ -250,17 +250,14 @@ You can access all lecture notes, videos and exercises online through the links ---> - - - - diff --git a/content/teaching/bmlip.md b/content/teaching/bmlip.md index ecdf6f4..6281118 100644 --- a/content/teaching/bmlip.md +++ b/content/teaching/bmlip.md @@ -29,7 +29,7 @@ Note: This site is currently under construction. ---> -The 2023/24 course "Bayesian Machine Learning and Information Processing" starts in November 2023 (Q2). +The 2023/24 course "Bayesian Machine Learning and Information Processing" starts in November 2023 (Q2). @@ -45,16 +45,16 @@ This course covers the fundamentals of a Bayesian (i.e., probabilistic) approach News and Announcements -- (20-Dec-2023) There is an issue with viewing the lecture notes. Please see [this Piazza note](https://piazza.com/class/ln1m8n333g5aa/post/79) for a temporary solution. +- (20-Dec-2023) There is an issue with viewing the lecture notes. Please see [this Piazza note](https://piazza.com/class/ln1m8n333g5aa/post/79) for a temporary solution. -- (15-Nov-2023) Please sign up for Piazza (Q&A platform) at [signup link](https://piazza.com/tue.nl/winter2024/5ssd0). As much as possible we will use the Piazza site for new announcements as well. +- (15-Nov-2023) Please sign up for Piazza (Q&A platform) at [signup link](https://piazza.com/tue.nl/winter2024/5ssd0). As much as possible we will use the Piazza site for new announcements as well. ## Instructors - [Prof.dr.ir. Bert de Vries](http://bertdv.nl) (email: bert.de.vries@tue.nl) is the responsible instructor for this course and teaches the [lectures with label B](#lectures). -- [Dr. Wouter Kouw](https://biaslab.github.io/member/wouter/) (w.m.kouw@tue.nl) teaches the probabilistic programming [lectures with label W](#lectures). -- [Tim Nisslbeck](https://biaslab.github.io/member/tim), [Sepideh Adamiat](https://biaslab.github.io/member/sepideh) and [Wouter Nuijten](https://biaslab.github.io/member/woutern) are the teaching assistants. +- [Dr. Wouter Kouw](https://biaslab.github.io/member/wouter/) (w.m.kouw@tue.nl) teaches the probabilistic programming [lectures with label W](#lectures). +- [Tim Nisslbeck](https://biaslab.github.io/member/tim), [Sepideh Adamiat](https://biaslab.github.io/member/sepideh) and [Wouter Nuijten](https://biaslab.github.io/member/woutern) are the teaching assistants. ## Materials @@ -62,9 +62,9 @@ In principle, you can download all needed materials from the links below. ### Lecture Notes -The lecture notes are mandatory material for the exam: +The lecture notes are mandatory material for the exam: -- Bert de Vries (2023), [PDF bundle of all lecture notes for lessons B0 through B12](https://github.com/bertdv/BMLIP/blob/master/output/BMLIP-5SSD0-lecture-notes.pdf). +- Bert de Vries (2023), [PDF bundle of all lecture notes for lessons B0 through B12](https://github.com/bertdv/BMLIP/blob/master/output/BMLIP-5SSD0-lecture-notes.pdf). - Wouter Kouw (2023), [PDF bundle of all probabilistic programming lecture notes for lessons W1 through W4](https://github.com/bertdv/BMLIP/blob/master/output/BMLIP-5SSD0-prob-prog.pdf). @@ -81,7 +81,7 @@ Learning](https://www.microsoft.com/en-us/research/people/cmbishop/prml-book/). ### Software - Please install Microsoft's VS Code editor ([download](https://code.visualstudio.com/Download)) and add the Jupyter notebook extension ([tutorial](https://code.visualstudio.com/learn/educators/notebooks)). -- Please install Julia version 1.9 ([download](https://julialang.org/downloads/)) on your machine and then add the Julia extension in VS Code ([tutorial](https://code.visualstudio.com/docs/languages/julia)). +- Please install Julia version 1.9 ([download](https://julialang.org/downloads/)) on your machine and then add the Julia extension in VS Code ([tutorial](https://code.visualstudio.com/docs/languages/julia)). ### Lecture notes, exercises, assignment and video recordings @@ -285,11 +285,11 @@ You can access all lecture notes, videos and exercises online through the links ### programming assignments - Programming assignments can be downloaded from the table above. -- Programming assignments should be submitted before the indicated deadlines at the [Canvas Assignments tab](https://canvas.tue.nl/courses/26086/assignments). +- Programming assignments should be submitted before the indicated deadlines at the [Canvas Assignments tab](https://canvas.tue.nl/courses/26086/assignments). ### grading -- See [this Piazza note on how the final grade is computed](https://piazza.com/class/ln1m8n333g5aa/post/6). +- See [this Piazza note on how the final grade is computed](https://piazza.com/class/ln1m8n333g5aa/post/6). - - - - diff --git a/layouts/partials/banner.html b/layouts/partials/banner.html index 0980e10..7cebaed 100644 --- a/layouts/partials/banner.html +++ b/layouts/partials/banner.html @@ -137,7 +137,7 @@ .st134{fill:#A3A3A3;} - + @@ -208,9 +208,9 @@ c0,0-181.4-1.6-314.9-2.3c-133.9-0.7-213.5-2.1-307.2-5.2c-90.5-3-248.2-6.6-254.6-8.9S3125.7,1822.9,3125.7,1822.9z"/> - - @@ -1973,7 +1973,7 @@ - @@ -1982,15 +1982,15 @@ - - + - @@ -2017,13 +2017,13 @@ - - @@ -2057,7 +2057,7 @@ C2382.3,894.5,2381.2,895.7,2381.2,897.2z"/> - @@ -2071,14 +2071,14 @@ - - @@ -2086,7 +2086,7 @@ C2253.5,943.7,2252.4,942.5,2250.9,942.5C2250.9,942.5,2250.9,942.5,2250.9,942.5z"/> - @@ -2094,7 +2094,7 @@ C2319.5,845.4,2320.6,846.5,2322.1,846.5z"/> - @@ -2164,7 +2164,7 @@ - + @@ -2216,7 +2216,7 @@ C2251.1,1118.9,2250,1117.7,2248.5,1117.6z"/> - @@ -2250,7 +2250,7 @@ C2252.4,1048.8,2250.8,1048.8,2249.7,1049.8C2249.7,1049.8,2249.7,1049.8,2249.7,1049.8z"/> - @@ -2258,7 +2258,7 @@ C2248.5,1348.1,2248.5,1349.7,2249.6,1350.8C2249.6,1350.7,2249.6,1350.7,2249.6,1350.8z"/> - @@ -2269,7 +2269,7 @@ - + @@ -2297,7 +2297,7 @@ - + @@ -2322,14 +2322,14 @@ C2381.3,1188.1,2380.2,1189.3,2380.2,1190.8z"/> - - @@ -2337,7 +2337,7 @@ C2298.2,1296.8,2297.1,1295.7,2295.6,1295.7z"/> - @@ -2345,7 +2345,7 @@ C2364.2,1198.4,2365.3,1199.6,2366.8,1199.6z"/> - @@ -2407,7 +2407,7 @@ - @@ -2415,7 +2415,7 @@ C2472.2,763.5,2472.2,761.9,2471.2,760.8z"/> - @@ -2424,7 +2424,7 @@ - + @@ -2444,7 +2444,7 @@ C2413.2,857.8,2414.4,856.6,2414.4,855.2z"/> - @@ -2471,7 +2471,7 @@ C2518.9,761.2,2520.1,760.1,2520.1,758.6z"/> - @@ -2557,7 +2557,7 @@ - + @@ -2566,7 +2566,7 @@ - + @@ -2584,26 +2584,26 @@ C2495.9,762.8,2497.1,761.6,2497.1,760.2z"/> - - - - @@ -2620,7 +2620,7 @@ - @@ -2645,7 +2645,7 @@ - @@ -2698,7 +2698,7 @@ C2501.5,1110.5,2502.6,1109.3,2502.7,1107.9z"/> - @@ -2706,7 +2706,7 @@ C2515.1,1116.8,2515.1,1115.2,2514.1,1114.1C2514.1,1114.1,2514.1,1114.1,2514.1,1114.1z"/> - @@ -2726,7 +2726,7 @@ - @@ -2764,7 +2764,7 @@ - @@ -2778,25 +2778,25 @@ C2419,1325.3,2417.8,1326.4,2417.8,1327.9z"/> - - - - @@ -2823,13 +2823,13 @@ C2503.9,1394.1,2505.1,1392.9,2505.1,1391.4z"/> - - @@ -2981,63 +2981,63 @@ s-171,192.6-244,318c-66.6,114.5-147.3,293.4-175.1,375.8c-21,62.2-39.4,153.4-41.8,162.8s252.9,32.2,492.7,14.9 c233.6-16.9,402-47.7,402-47.7S2000.7,1725.1,1975.7,1641.1z"/> - - - - - - - - - - - - @@ -3046,29 +3046,29 @@ - - - - - - - - @@ -3201,13 +3201,13 @@ - - - @@ -3216,7 +3216,7 @@ - @@ -3530,7 +3530,7 @@ c0-0.4,0-0.8-0.1-1.2l8.2-1.4C3919.4,1681.1,3919,1683,3918.4,1685.2L3918.4,1685.2z"/> - + @@ -3545,7 +3545,7 @@ - + @@ -3576,63 +3576,63 @@ s-171,192.6-244,318c-66.6,114.5-147.3,293.4-175.1,375.8c-21,62.2-39.4,153.4-41.8,162.8s252.9,32.2,492.7,14.9 c233.6-16.9,402-47.7,402-47.7S2000.7,1725.3,1975.7,1641.3z"/> - - - - - - - - - - - - @@ -3641,29 +3641,29 @@ - - - - - - - - diff --git a/layouts/partials/index/contact.html b/layouts/partials/index/contact.html index 37730a4..e2cdcea 100644 --- a/layouts/partials/index/contact.html +++ b/layouts/partials/index/contact.html @@ -1,5 +1,5 @@
    - + {{ with .Site.Params.youtube }}
  • diff --git a/layouts/partials/slogan.html b/layouts/partials/slogan.html index b31bff6..0fc7dd7 100644 --- a/layouts/partials/slogan.html +++ b/layouts/partials/slogan.html @@ -20,4 +20,4 @@
    {{partial "banner.html" .}}
    - \ No newline at end of file + diff --git a/layouts/section/open-projects.html b/layouts/section/open-projects.html index 129c57f..3ce95b5 100644 --- a/layouts/section/open-projects.html +++ b/layouts/section/open-projects.html @@ -9,24 +9,24 @@

    Open projects

    {{"We have several open research projects for MSc students who want to do an internship or graduation project. We keep these projects in a portfolio document: [BIASlab MSc Project Portfolio](/pdf/msc-project-portfolio.pdf). These are either in collaboration with a company (good for your professional experience) or are purely for internal academic research (if you prefer to have more freedom). In general, the projects focus on signal processing applications, control systems applications or probabilistic programming software. For examples of these categories, see the list below:" | markdownify}}

    - + {{ range .Data.Pages.ByDate.Reverse }}
    • - + {{ if .Content }} {{ .Scratch.Set "link" .Permalink }} {{ else if .Params.external_link }} {{ .Scratch.Set "link" .Params.external_link }} {{ end }} - +

      {{ with .Params.description }} {{ . }} {{ end }} - +
    @@ -37,7 +37,7 @@

    {{"If you have your own ideas for a neuro-inspired AI / probabilistic machine learning MSc project, feel free to contact prof. [Bert de Vries](http://bertdv.nl)." | markdownify }} - +
    diff --git a/static/css/biaslab.css b/static/css/biaslab.css index 0112693..902d4c0 100644 --- a/static/css/biaslab.css +++ b/static/css/biaslab.css @@ -149,7 +149,7 @@ header{ svg:not(:root){ position: relative; - left: 0; + left: 0; top: 0; } @@ -175,4 +175,4 @@ a.open-positions { content: attr(data-ribbon); color: #fff; font: 400 1.5em "Lato", Helvetica, Arial, sans-serif; -} \ No newline at end of file +} diff --git a/static/fonts/biaslab-font.svg b/static/fonts/biaslab-font.svg index 6ab6f42..70c97bc 100644 --- a/static/fonts/biaslab-font.svg +++ b/static/fonts/biaslab-font.svg @@ -9,4 +9,4 @@ - \ No newline at end of file +