Path Structured Multimarginal Schrödinger Bridge for Probabilistic Learning of Hardware Resource Usage by Control Software
This provides a probabilistic learning approach for predicting cyber-physical context-dependent performance in software, which is incremental as it applies existing algorithmic advances to a new domain.
The paper tackled the problem of predicting hardware resource usage by control software by solving a path structured multimarginal Schrödinger bridge problem, resulting in a method that predicts time-varying distributions with guaranteed linear convergence and rapid convergence to accurate predictions.
The solution of the path structured multimarginal Schrödinger bridge problem (MSBP) is the most-likely measure-valued trajectory consistent with a sequence of observed probability measures or distributional snapshots. We leverage recent algorithmic advances in solving such structured MSBPs for learning stochastic hardware resource usage by control software. The solution enables predicting the time-varying distribution of hardware resource availability at a desired time with guaranteed linear convergence. We demonstrate the efficacy of our probabilistic learning approach in a model predictive control software execution case study. The method exhibits rapid convergence to an accurate prediction of hardware resource utilization of the controller. The method can be broadly applied to any software to predict cyber-physical context-dependent performance at arbitrary time.