Skip to content

Latest commit

 

History

History
76 lines (52 loc) · 2.5 KB

README.md

File metadata and controls

76 lines (52 loc) · 2.5 KB
Raw

AutoEncoders in PyTorch

dep2 dep1

Description

This repo contains an implementation of the following AutoEncoders:

  • Vanilla AutoEncoders - AE:
    The most basic autoencoder structure is one which simply maps input data-points through a bottleneck layer whose dimensionality is smaller than the input.

  • Variational AutoEncoders - VAE:
    The Variational Autoencoder introduces the constraint that the latent code z is a random variable distributed according to a prior distribution p(z).

  • Adversarially Constrained Autoencoder Interpolations - ACAI:
    A critic network tries to predict the interpolation coefficient α corresponding to an interpolated datapoint. The autoencoder is trained to fool the critic into outputting α = 0.
    ACAI-figure

Setup

Create a Python Virtual Environment

mkvirtualenv --python=/usr/bin/python3 pytorch-AE

Install dependencies

pip install torch torchvision

Training

python train.py --help

Training Options and some examples:

  • Vanilla Autoencoder:

    python train.py --model AE

  • Variational Autoencoder:

    python train.py --model VAE --batch-size 512 --dataset EMNIST --seed 42 --log-interval 500 --epochs 5 --embedding-size 128

Results

Vanilla AutoEncoders Variational AutoEncoders ACAI

Contributing:

If you have suggestions or any type of contribution idea, file an issue, make a PR and don't forget to star the repository

More projects:

Feel free to check out my other repos with more work in Machine Learning: