English

Periodic Extrapolative Generalisation in Neural Networks

Machine Learning 2022-09-22 v1 Artificial Intelligence

Abstract

The learning of the simplest possible computational pattern -- periodicity -- is an open problem in the research of strong generalisation in neural networks. We formalise the problem of extrapolative generalisation for periodic signals and systematically investigate the generalisation abilities of classical, population-based, and recently proposed periodic architectures on a set of benchmarking tasks. We find that periodic and "snake" activation functions consistently fail at periodic extrapolation, regardless of the trainability of their periodicity parameters. Further, our results show that traditional sequential models still outperform the novel architectures designed specifically for extrapolation, and that these are in turn trumped by population-based training. We make our benchmarking and evaluation toolkit, PerKit, available and easily accessible to facilitate future work in the area.

Keywords

Cite

@article{arxiv.2209.10280,
  title  = {Periodic Extrapolative Generalisation in Neural Networks},
  author = {Peter Belcák and Roger Wattenhofer},
  journal= {arXiv preprint arXiv:2209.10280},
  year   = {2022}
}

Comments

Accepted to IEEE Symposium on Deep Learning (IEEE DL) 2022, 8 pages, 7 figures

R2 v1 2026-06-28T01:48:35.017Z