The project is a projectory of my theorethical knowledge of the foundations of Neural Networks and how they work behind the scenes.

• Every layer, loss or activation function, optimization algorithm, gradient computation, and model in general - was made with no libraries implementation uses. The Whole math was done by hand and executed using Numpy arrays for convinience only!

• All implementations are organised in the code in self-explanatory names, and the "Tests.py" script contain many functions to demonstrate the many phases completed in the project, aswell as tests (such as implementing and utilizing Gradient&Jacobian Checks) to make sure the math and computations are in check.

• You may also find many "non-used" functions such as SGD variations that are there to give a sneak-pic of my experimentations along the writing of the project.

• The project revolves around creating a classification model for a syntethic set of data with various input dimensions and output, with a CrossEntropy loss function, and supportage of both classic FC layers and Residual FC layers in a mode.

• Enjoy reviewing and playing with the code yourself!