作者
Ferran Alet, Dylan Doblar, Allan Zhou, Josh Tenenbaum, Kenji Kawaguchi, Chelsea Finn
发表日期
2021/12/6
期刊
Advances in Neural Information Processing Systems
卷号
34
页码范围
16384-16397
简介
Progress in machine learning (ML) stems from a combination of data availability, computational resources, and an appropriate encoding of inductive biases. Useful biases often exploit symmetries in the prediction problem, such as convolutional networks relying on translation equivariance. Automatically discovering these useful symmetries holds the potential to greatly improve the performance of ML systems, but still remains a challenge. In this work, we focus on sequential prediction problems and take inspiration from Noether's theorem to reduce the problem of finding inductive biases to meta-learning useful conserved quantities. We propose Noether Networks: a new type of architecture where a meta-learned conservation loss is optimized inside the prediction function. We show, theoretically and experimentally, that Noether Networks improve prediction quality, providing a general framework for discovering inductive biases in sequential problems.
学术搜索中的文章
F Alet, D Doblar, A Zhou, J Tenenbaum, K Kawaguchi… - Advances in Neural Information Processing Systems, 2021