Generative Fourier-based Auto-Encoders

This is a PyTorch implementation for the framework presented in the following paper: Generative Fourier-based Auto-Encoders: Preliminary Results paper

Requirements

• torch==1.5.0
• torchaudio==0.5.0
• scikit-learn==0.22.2.post1
• numpy==1.12.0

Use requirements.txt to install all the dependencies. Tested only with Python 3.6

Data

Download the dataset used from here and place it in data.

Run the experiment

• check if all the path are corrects in the settings of the various scripts
• to test the likelihood of the models run likelihood/run_likelihood.py
• to obtain the Fig.2a use likelihood/print_likelihood.py
• with the number of components obtain from the previous script, run ppca/experiment_ppca.py
• with the data saved from the previous script, run ppca/plot_generation.py to obtain Fig.2b