jiristepanovsky.cz

Return to Index

Deep Neural Network - 2021

This report contains my implementation of a Deep Neural Network with Gradient descent learning. The report explains the mathematical theory neural networks are based on and demonstrates its successful use in a curve-fitting task and MNIST dataset image classification. The implementation is comparable to Keras deep learning library in terms of accuracy but is about 3 times slower, which is to be expected for pure python implementation.

The code, written in Jupyter Notebook, can be downloaded here.
The report (below) can be downloaded here.