Probabilistic Testing of Probabilistic Programs

This is a project in computer science and software
engineering, with a hint of applied Bayesian statistics.
The goal is to build testing theory and infrastructure
for testing probabilistic programs. Testing is the
leading quality assurance method for software.
Randomness is increasingly used in software
(especially in machine learning, statistics, and
robotics). Unfortunately, randomness and testing do
not mix well, as randomness makes test results flaky -
the test outcome is not deterministically the same for
each run. Thus, we need a new testing theory and
method for probabilistic programs. The goal of the
project is to understand 1) What kind of properties
are testable for probabilistic programs? 2) how to test
these properties in practice, so that the test are
reliable, so avoiding false positives and false
negatives?