Elempereur / WCRG

Code for paper "Conditionally Strongly Log-Concave Generative Model"

Wavelet-Conditional-Renormalization-Group

Code for paper "Conditionally Strongly Log-Concave Generative Model".

This repository implements fast wavelet transform and wavelet packets transform in pytorch. As well, it implements the wavelet conditional renormalisation group with score matching.

Install

For fast wavelet transform and wavelet packets transform in pytorch download the folder Wavelet_Packets :

import sys
sys.path.append('~/where/you/download/the/script/')

Then, import it:

import Wavelet_Packets

For wavelet conditional renormalisation group, as it is a package that uses Wavelet_Packets, you must download both folders Wavelet_Packets and WCRG

import sys
sys.path.append('~/where/you/download/the/script/')

Then, import both:

import Wavelet_Packets
import WCRG

Running Examples

The folder Notebooks Examples contains a tutorial for the use of wavelet packets and fast wavelet transform. As well, an example of how to use the model to learn energies, free energies and sample, for $\varphi^4$ at critical point.

Using the software

You are free to use this software for academic purposes only. If you do so, please cite paper "Conditionally Strongly Log-Concave Generative Model". Do not hesitate to message me if you have any question.