Train and analyze different types of models for semi-supervised learning. Contains implementations for logistic regression, linear VAE, convolutional VAE, M1 model and Prediction-constrained VAE. Currently supports MNIST and CIFAR-10 datasets.
Example: python3 scripts/train_vae.py --config scripts/configs/config-vae.json
You should run the scripts from the root directory (semisupervisedVAE). You may have to run this command first:
export PYTHONPATH="/semisupervisedVAE:$PYTHONPATH"
You can install requirements using pip install -r requirements.txt --no-cache-dir directly or in a virtual environment.