Deep Learning Introduction

This repository contains a LaTeX presentation on the fundamentals of neural networks, including notation, backpropagation algorithm, stochastic gradient descent, and practical examples.

The PDF with the latest version of the slides can be found here.

Examples Covered

• Binary classifier on 2D feature space
• Convolutional neural network on MNIST dataset

Figures

• 2D classification dataset

• Training animation video

training_animation.mp4

Compilation

For $\LaTeX$ compilation, it is necessary to generate the Julia lexer for pymentize for the minted environment. For this I am using sisl/pygments-julia within a virtual environment. Using uv, you can install the necessary Pygments lexer by running

uv pip install --system git+https://github.com/sisl/pygments-julia#egg=pygments_julia 

Then check that julia1 is available as a pygmentize lexer with

pygmentize -L lexers | rg julia1

Note

You must also have the Pygments python package installed. Follow the instructions in the documentation to install the package.

You must ensure that the directory for running the pygmentize external command exists. This can be set to run in the base directory in Tectonic.toml or you must create the directory shell-escape manually with

mkdir shell-escape

Finally the slides can be compiled with tectonic

tectonic -X build

Note

You can also install the necessary python packages within a virtual environment. You can use uv for this by running uv venv and then installing the packages without the --system flag. Keep in mind that the virtual environment must be active for the compilation to work.