ANC Workshop - Nigel Goddard, Cian Eastwood

Tuesday, 9th May 2023

Exploiting the IDEAL home energy dataset - Nigel Goddard

Abstract: IDEAL was a research project that ran from 2013 to 2017, which gathered a uniquely extensive and fine-grained dataset on energy use in 255 homes.  This talk will present 3 recent student studies based on the IDEAL and related datasets, and discuss some future plans for further studies, each relavent to the energy transition or social care.

Probable domain generalization via quantile risk minimization - Cian Eastwood

Abstract: Domain generalization (DG) seeks predictors which perform well on unseen test distributions by leveraging data drawn from multiple related training distributions or domains. To achieve this, DG is commonly formulated as an average- or worst-case problem over the set of possible domains. However, predictors that perform well on average lack robustness while predictors that perform well in the worst case tend to be overly conservative. To address this, we propose a new probabilistic framework for DG where the goal is to learn predictors that perform well _with high probability_. Our key idea is that distribution shifts seen during training should inform us of probable shifts at test time, which we realize by explicitly relating training and test domains as draws from the same underlying meta-distribution. Then, by minimizing a particular quantile of predictors' performance distributions over training domains, we learn predictors that perform well on unseen test domains with the corresponding probability.

Event type: Workshop

Date: Tuesday, 9th May 2023

Time: 11:00

Location: G.03

Speaker(s): Nigel Goddard, Cian Eastwood

Chair/Host: Sigrid Hellan