dangbert / deeplearningpython35

DeepLearningPython35

Purpose

This is a learning project for developing neural networks from scratch in python. This is my work as I follow the neuralnetworksanddeeplearning.com book guide.

General Goal: Make it through the book, then later consider greateer code deviations from the book (or start a new branch/repo)/

• This repository is a fork of MichalDanielDobrzanski/DeepLearningPython35 which is itself a python 3.5 port of the book's provided code repository.

Overview

Provided Files:

Maybe slightly modified from initial provided code in repo.

• test.py:
  • File for running (training/evaluating) all networks provided with the initial code repo (see below). (Note: I haven't made any substantial changes).
• network.py: Neural net implementation provided (with initial repo) for chapters 1, 2.
• network2.py: Neural net implementation provided for chapters 3, 4, 5.
• network3.py: Neural net implementation provided for chapter 6 (the last chapter).
• mnist files:
  • mnist.pkl.gz: mnist data set
  • expand_mnist.py
  • mnist_average_darkness.py
  • mnist_loader.py
  • mnist_svm.py

My Personal Files:

• test1.py: My test runner for mynet.py
• mynet.py: My neural net implementation based on network.py
• learning-network.ipynb: Jupyter Notebook for experimenting
• bootyNet.py: incomplete personal idea not directly related to this book

Run unit tests:

# setup virtualenv (if not done previously):
virtualenv env
source env/bin/activate
pip3 install -r requirements.txt
pytest -s

Build docs:

source env/bin/activate
cd docs
# build html docs:
make html
# build pdf docs:
sudo apt-install latexmk texlive-latex-recommended texlive-latex-extra
make latexpdf

---

Future Ideas:

• Generate adversarial images
• Implement GAN's?
• Revisit bootyNet.py and look over my neural net ideas list, notion list
• Do some cool experiments with visualizing neurons in my network:
  • https://distill.pub/2017/feature-visualization/
  • https://distill.pub/2020/circuits/zoom-in/
  • manim library
    • Getting Started Tutorial