Rare event modeling with self-regularized normalizing flows: what can we learn from a single failure?

Increased deployment of autonomous systems in fields like transportation and robotics has led to a corresponding increase in safety-critical failures. These failures are difficult to model and debug due to the relative lack of data: while normal operations provide tens of thousands of examples, we may have only seconds of data leading up to the failure. This scarcity makes it challenging to train generative models of rare failure events, as existing methods risk either overfitting to noise in the limited failure dataset or underfitting due to an overly strong prior. We address this challenge with CALNF, or calibrated normalizing flows, a selfregularized framework for posterior learning from limited data. CALNF achieves state-of-the-art performance on data-limited failure modeling problems and enables a first-of-a-kind case study of the 2022 Southwest Airlines scheduling crisis.