Title: Approximate Bayesian Computation with Path Signatures
Short description: (Winner of the Outstanding Paper Award at UAI 2024) Approximate Bayesian computation (ABC) generates likelihood-free posterior samples by comparing simulated and observed data through some distance measure. Existing approaches to ABC are often poorly suited to time series simulators, for example due to an independent and identically distributed data assumption. In this paper, we propose to use path signatures in ABC to handle the sequential nature of time series. We provide theoretical guarantees on the resultant posteriors and demonstrate competitive Bayesian parameter inference for simulators generating univariate, multivariate, and irregularly spaced sequences of non-iid data.
Title: Approximate Bayesian Computation with Path Signatures
Short description: (Winner of the Outstanding Paper Award at UAI 2024) Approximate Bayesian computation (ABC) generates likelihood-free posterior samples by comparing simulated and observed data through some distance measure. Existing approaches to ABC are often poorly suited to time series simulators, for example due to an independent and identically distributed data assumption. In this paper, we propose to use path signatures in ABC to handle the sequential nature of time series. We provide theoretical guarantees on the resultant posteriors and demonstrate competitive Bayesian parameter inference for simulators generating univariate, multivariate, and irregularly spaced sequences of non-iid data.
Link to paper: https://openreview.net/forum?id=MsU6sKOcA4#