KQE: Kernel Quantile Embeddings and Discrepancies

A Python library for computing Kernel Quantile Discrepancies (KQDs) based on Kernel Quantile Embeddings (KQEs), proposed in PAPER.

Features

• Main module: kqd
• Backends: Built on JAX for accelerated, differentiable computations.
• Experiments: Experiments from the PAPER are implemented in the experiments/ folder.

Installation

1. Clone the repository

git clone https://github.com/mashanaslidnyk/kqe.git
cd kqe

2. Install the package

• Editable mode (recommended for development):

  make dev-install

• Editable mode that also allows you run experiments:

  To install in editable mode, and also install dependencies required to run the experiments in the experiments/ folder:

  make dev-install-experiments

• Standard install:

  make install

Quick Start

import jax.numpy as jnp
from kqe.kernels import GaussianKernel
from kqe.kqd import ekqd

# Sample data
X = jnp.array([[1.0], [2.0], [3.0]])
Y = jnp.array([[1.5], [2.5], [3.5]])

# Kernel
k = GaussianKernel(l=1.0)

# Compute e-KQD²
ekqd_val = ekqd(X, Y, kernel_fn=k, num_projections=3)
print("eKQD²:", ekqd_val)

Testing & Formatting

To run pytest tests:

make test

To format everything with black and isort:

make format

License

This project is licensed under the GNU GPLv3. See the LICENSE file for details.