AutoEncoders

A small, growing notebook series that walks through different types of autoencoders. Current notebook focuses on a convolutional autoencoder on MNIST. Future notebooks will cover variants like VAE, denoising, and sparse autoencoders.

Contents

• autoencoder.ipynb: convolutional autoencoder on MNIST with reconstruction and latent PCA visualization.

Setup (conda)

conda env create -f environment.yml
conda activate autoencoders

Setup (pip)

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

Run

jupyter lab

Open autoencoder.ipynb and run all cells in order.

Notes

• MNIST downloads into ./data the first time you run the notebook.
• GPU is optional. If you want GPU support, install a matching PyTorch build for your system.