Manifold Mixup: Better Representations by Interpolating Hidden States

Vikas Verma, Alex Lamb, Christopher Beckham, Amir Najafi, Ioannis Mitliagkas, David Lopez-Paz, Yoshua Bengio

Tutkimustuotos: Artikkeli kirjassa/konferenssijulkaisussaConference article in proceedingsScientificvertaisarvioitu

388 Lataukset (Pure)


Deep neural networks excel at learning the training data, but often provide incorrect and confident predictions when evaluated on slightly different test examples. This includes distribution shifts, outliers, and adversarial examples. To address these issues, we propose Manifold Mixup, a simple regularizer that encourages neural networks to predict less confidently on interpolations of hidden representations. Manifold Mixup leverages semantic interpolations as additional training signal, obtaining neural networks with smoother decision boundaries at multiple levels of representation. As a result, neural networks trained with Manifold Mixup learn class-representations with fewer directions of variance. We prove theory on why this flattening happens under ideal conditions, validate it on practical situations, and connect it to previous works on information theory and generalization. In spite of incurring no significant computation and being implemented in a few lines of code, Manifold Mixup improves strong baselines in supervised learning, robustness to single-step adversarial attacks, and test log-likelihood.
OtsikkoProceedings of the 36th International Conference on Machine Learning
TilaJulkaistu - 2019
OKM-julkaisutyyppiA4 Artikkeli konferenssijulkaisussa
TapahtumaInternational Conference on Machine Learning - Long Beach, Yhdysvallat
Kesto: 9 kesäk. 201915 kesäk. 2019
Konferenssinumero: 36


NimiProceedings of Machine Learning Research
ISSN (elektroninen)6438-6447


ConferenceInternational Conference on Machine Learning
KaupunkiLong Beach


Sukella tutkimusaiheisiin 'Manifold Mixup: Better Representations by Interpolating Hidden States'. Ne muodostavat yhdessä ainutlaatuisen sormenjäljen.

Siteeraa tätä