variationally-weighted-kernel-density-estimation

This repository contains a PyTorch implementation for the paper Variational Weighting for Kernel Density Ratios, presented at NeurIPS 2023.

Dependencies

• PyTorch

• Scikit-learn

• SciPy

• NumPy

• CuPy

Running the experiments

python train.py --model [model name] --device [device]

Here model name is one of the following:

• KDE This is the naive Kernel Density Estimation model.
• based Corresponds to VWKDE-MB in Figure 4.
• free Corresponds to VWKDE-MF in Figure 4.

Example:

python train.py --model free --device cuda