VMC: A Dafny Library for Verified Monte Carlo Algorithms

VMC: A Dafny Library for Verified Monte Carlo Algorithms

Explore a conference talk introducing VMC, an open-source Dafny library for verified randomized algorithms of the Monte Carlo type. Delve into the library's offerings of verified samplers for standard distributions, applicable in Dafny, C#, and Java code. Examine the three-part structure of each sampler: a functional model transforming bitstreams to distribution domain values, a proof validating the model's distribution, and an imperative implementation using external random bits, proven equivalent to the functional model. Follow a detailed walkthrough of each step using the uniform distribution as an example. Gain insights into the project's current status and potential future developments in the field of verified Monte Carlo algorithms.