Project: Interventional Whittle Sum-Product Networks
Description
Whittle sum-product networks [1] model the joint distribution of multivariate time series by leveraging the Whittle approximation, casting the likelihood in the frequency domain, and place a complex-valued sum-product network over the frequencies. The conditional independence relations among the time series can then be determined efficiently in the spectral domain. Granger causality [2] is the framework that deals with causal effects in the time series domain. In this project we will develop an interventional Whittle SPN that can effectively handle interventions in the time series.
References:
[1] Yu et al., Whittle Networks: A Deep Likelihood Model for Time Series, ICML 2021
[2] Eichler and Didelez, On Granger causality and the effect of interventions in time series, Lifetime Data Analysis 2009
[3] Zečević et al., Interventional Sum-Product Networks: Causal Inference with Tractable Probabilistic Models, NeurIPS 2021
[4] Runge et al., Causal inference for time series, Nature reviews Earth & Environment 2023
Details
- Supervisor
-
Devendra Dhami
- Interested?
- Get in contact