Abstract
Neural networks offer good approximation to many tasks but consistently fail to reach perfect generalization, even when theoretical work shows that such perfect solutions can be expressed by certain architectures. Using the task of formal language learning, we focus on one simple formal language and show that the theoretically correct solution is in fact not an optimum of commonly used objectives — even with regularization techniques that according to common wisdom should lead to simple weights and good generalization (L1, L2) or other meta-heuristics (early-stopping, dropout). On the other hand, replacing standard targets with the Minimum Description Length objective (MDL) results in the correct solution being an optimum.- Anthology ID:
- 2024.acl-long.713
- Volume:
- Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
- Month:
- August
- Year:
- 2024
- Address:
- Bangkok, Thailand
- Editors:
- Lun-Wei Ku, Andre Martins, Vivek Srikumar
- Venue:
- ACL
- SIG:
- Publisher:
- Association for Computational Linguistics
- Note:
- Pages:
- 13198–13210
- Language:
- URL:
- https://aclanthology.org/2024.acl-long.713
- DOI:
- Cite (ACL):
- Nur Lan, Emmanuel Chemla, and Roni Katzir. 2024. Bridging the Empirical-Theoretical Gap in Neural Network Formal Language Learning Using Minimum Description Length. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 13198–13210, Bangkok, Thailand. Association for Computational Linguistics.
- Cite (Informal):
- Bridging the Empirical-Theoretical Gap in Neural Network Formal Language Learning Using Minimum Description Length (Lan et al., ACL 2024)
- PDF:
- https://preview.aclanthology.org/nschneid-patch-4/2024.acl-long.713.pdf