Loss Surface Simplexes for Mode Connecting Volumes and Fast Ensembling

Gregory Benton · Wesley Maddox · Sanae Lotfi · Andrew Wilson

Keywords: [ Computer Vision ] [ Deep Learning ] [ Applications ] [ Supervised Deep Networks ] [ Algorithms -> Classification; Deep Learning; Deep Learning ]


With a better understanding of the loss surfaces for multilayer networks, we can build more robust and accurate training procedures. Recently it was discovered that independently trained SGD solutions can be connected along one-dimensional paths of near-constant training loss. In this paper, we in fact demonstrate the existence of mode-connecting simplicial complexes that form multi-dimensional manifolds of low loss, connecting many independently trained models. Building on this discovery, we show how to efficiently construct simplicial complexes for fast ensembling, outperforming independently trained deep ensembles in accuracy, calibration, and robustness to dataset shift. Notably, our approach is easy to apply and only requires a few training epochs to discover a low-loss simplex.

Chat is not available.