On Scalable Testing of Samplers

In this paper we study the problem of testing of constrained samplers over high-dimensional distributions with (varepsilon,η,δ) guarantees. Samplers are increasingly used in a wide range of safety-critical ML applications, and hence the testing problem has gained importance. For n-dimensional distributions, the existing state-of-the-art algorithm, mathsfBarbarik2, has a worst case query complexity of exponential in n and hence is not ideal for use in practice. Our primary contribution is an exponentially faster algorithm that has a query complexity linear in n and hence can easily scale to larger instances. We demonstrate our claim by implementing our algorithm and then comparing it against mathsfBarbarik2. Our experiments on the samplers mathsfwUnigen3 and mathsfwSTS, find that mathsfBarbarik3 requires 10× fewer samples for mathsfwUnigen3 and 450× fewer samples for mathsfwSTS as compared to mathsfBarbarik2.