Efficiently Computable Safety Bounds for Gaussian Processes in Active Learning

arXiv:2402.18260v1 Announce Type: new
Abstract: Active learning of physical systems must commonly respect practical safety constraints, which restricts the exploration of the design space. Gaussian Processes (GPs) and their calibrated uncertainty estimations are widely used for this purpose. In many technical applications the design space is explored via continuous trajectories, along which the safety needs to be assessed. This is particularly challenging for strict safety requirements in GP methods, as it employs computationally expensive Monte-Carlo sampling of high quantiles. We …

abstract active learning applications arxiv constraints continuous cs.lg design estimations exploration gaussian processes gps practical processes safety space systems technical type uncertainty via