AI, Data Science & Cloud Certificates from Google, IBM & Meta — 40% Off
One plan covers every Professional Certificate on Coursera. 40% off Coursera Plus Annual.
Unlock All Certificates
Explore a conference presentation that introduces automatic tuning techniques for random generators in property-based testing through probabilistic programming. Learn how researchers from leading universities have developed methods to automatically optimize generator weights that determine the distribution of test inputs, addressing the traditionally tedious manual process of tuning these parameters. Discover the novel discrete probabilistic programming system called Loaded Dice, which supports differentiation and parameter learning for generator optimization. Understand how the approach enables users to target desired distributions and improve both diversity and validity of test cases through objective functions. Examine empirical results demonstrating 3.1-7.4x speedup in bug finding when generators are automatically tuned for diversity and validity across property-based testing benchmarks. Gain insights into how this work bridges property-based testing with probabilistic programming to make generator tuning more accessible and effective for software validation.
