mathieuchal / lft

The official repository for the paper "Learning Functional Transduction"

Learning Functional Transduction

This repository contains the code to replicate results from Learning Functional Transduction (Chalvidal, NeurIPS 2023). Specifically, it contains the scripts allowing to meta-learn solutions of the four presented experiments:

• Regression of 1D Advection-Diffusion-Reaction equations
• Regression of 2D Burgers systems with varying diffusion coefficients
• Prediction of climate variable under seasonal influence
• Classification of MNIST-like datasets

The code is written in Python and use Pytorch for dataset sampling and model definition. Data is accessible at the following (to come).