Abstract

Traditionally, only the last checkpoint of a training run is used for final model prediction, but so much more data is accumulated during training. In this talk, we will discuss ways of utilizing data from the entire training trajectory in the context of image classification to improve final model performance, at a fraction of the cost of traditional ensembles.

Louise Anderson-Conway

AI Resident @ Google

Louise Anderson-Conway is an AI resident at Google, where she researches model ensembles in the Perception group. She got her M.Sc. from Chalmers University in 2011, a Ph.D. in theoretical physics from Chalmers in 2015 and spent time as a postdoc at both Imperial College London and Stanford University before joining Google in 2020.