This is a python implementation of algorithms and experiemtns presented in the paper "Kernel Hypothesis Testing with Set-valued Data". The goal of this project is to provide a framework for hypothesis testing with data most appropriately described by individual distributions, and available to the researcher as sets of examples, such as time series. This file contains implementations of our tests and other alternatives, simple use cases of our tests, and performance comparisons for the two sample and independence problem.
The only significant dependencies are python 3.6 or later and tensorflow.
To get started, check our tutorials which will guide you through the tests from the beginning.
We acknowledge the use of the set-up and some kernel-based test from https://github.com/wittawatj/.