Loss Surface Simplexes for Mode Connecting Volumes and Fast Ensembling
This provides a more efficient way to build robust ensembles for machine learning practitioners, though it is incremental as it builds on prior work on one-dimensional mode connections.
The paper tackles the problem of connecting multiple independently trained neural network models via low-loss multi-dimensional manifolds, resulting in a fast ensembling method that outperforms deep ensembles in accuracy, calibration, and robustness to dataset shift with only a few training epochs.
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 show that there are mode-connecting simplicial complexes that form multi-dimensional manifolds of low loss, connecting many independently trained models. Inspired by this discovery, we show how to efficiently build simplicial complexes for fast ensembling, outperforming independently trained deep ensembles in accuracy, calibration, and robustness to dataset shift. Notably, our approach only requires a few training epochs to discover a low-loss simplex, starting from a pre-trained solution. Code is available at https://github.com/g-benton/loss-surface-simplexes.