Simulation-based optimisation of dynamic systems, concentrating on high-fidelity static and dynamic optimisation methods
Inference problems in the context of parametric (static) and control (dynamic) optimisation contain high-dimensional, richly-structured mechanistic models that only partially explain real outcomes of interval-valued time series data. This presents a key limitation on our ability to construct high-fidelity analytical and predictive optimisation models that can be combined efficiently with simulators in order to optimise for high-resolution complex, dynamical processes and behaviour underlying both stochastic and deterministic systems.
Improving generative models and simulation-based optimisation methods provides direct informational and methodological appeal to more efficiently understand, estimate, measure, and infer from real-world large-scale and complex stochastic processes. Toward this end, I’m investigating simulation-based dynamic programming and algorithms for predicting how systems behave across interval-valued time series models. This includes parametric and control optimisation methods for determining the values of parameters that maximise or minimise an objective function, and optimal or near-optimal control in a given state visited by a system; respectively.
Beyond conceptual and methodological considerations to characterise distributed control techniques; my preliminary technical objectives here are tethered more closely with establishing computationally tractable models and analysis methods that can reveal macroscopic behaviour and establish causality between interval data variables in discrete-event systems.