Language : MATLAB
This project is for non-iterative algorithm development which are mostly based on the least square method. All of these following algorithms are based on single-hidden layer feed-forward neural network (SLFN) structure. List of algorithms:
- ELM (Extreme Learning Machine), PCA-ELM (principal component analysis), I-ELM (incremental), EI-ELM (enhanced incremental), DP-ELM (destructive parsimonious), and CP-ELM (constructive parsimonious).
- AIL (Analitycal Incremental Learning)
- LSM (Local Sigmoid Method)
- BP (backpropagation, LM) matlab wraper.
How to run the comparison of all algorithms
- Regression case: run
runcvreg.m. - Classification case: run
runcvclass.m. - To find the hyperparameters: run
hpsearching.m, you can change the case to regression or classification. - To summary the metrics (accuracy, #nodes, time): run `perfsummay.m, you need to change the metrics manually --see/read the code.