Augmenting Classic Algorithms with Neural Components for Strong Generalisation on Ambiguous and High-Dimensional Data

Abstract

We augment classic algorithms with learned components to adapt them to domains currently dominated by deep learning models. Two traditional sorting algorithms with learnable neural building blocks are applied to visual data with apriori unknown symbols and rules. The models are quickly and reliably trained end-to-end in a supervised setting. Our models learn symbol representations and generalise better than generic neural network models to longer input sequences.

PDF NeurIPSW OpenReview Semantic Scholar

Cite

Text

Schlag and Schmidhuber. "Augmenting Classic Algorithms with Neural Components for Strong Generalisation on Ambiguous and High-Dimensional Data." NeurIPS 2021 Workshops: AIPLANS, 2021.

Markdown

[Schlag and Schmidhuber. "Augmenting Classic Algorithms with Neural Components for Strong Generalisation on Ambiguous and High-Dimensional Data." NeurIPS 2021 Workshops: AIPLANS, 2021.](https://mlanthology.org/neuripsw/2021/schlag2021neuripsw-augmenting/)

BibTeX

@inproceedings{schlag2021neuripsw-augmenting,
  title     = {{Augmenting Classic Algorithms with Neural Components for Strong Generalisation on Ambiguous and High-Dimensional Data}},
  author    = {Schlag, Imanol and Schmidhuber, Jürgen},
  booktitle = {NeurIPS 2021 Workshops: AIPLANS},
  year      = {2021},
  url       = {https://mlanthology.org/neuripsw/2021/schlag2021neuripsw-augmenting/}
}