A Congruence-based Approach to Active Automata Learning from Neural Language Models

Franz Mayr, Sergio Yovine, Matías Carrasco, Federico Pan, Federico Vilensky
Proceedings of 16th edition of the International Conference on Grammatical Inference, PMLR 217:250-264, 2023.

Abstract

The paper proposes an approach for probably approximately correct active learning of probabilistic automata (PDFA) from neural language models. It is based on a congruence over strings which is parameterized by an equivalence relation over probability distributions. The learning algorithm is implemented using a tree data structure of arbitrary (possibly unbounded) degree. The implementation is evaluated with several equivalences on LSTM and Transformer-based neural language models from different application domains.

Cite this Paper

BibTeX
@InProceedings{pmlr-v217-mayr23a, title = {A Congruence-based Approach to Active Automata Learning from Neural Language Models}, author = {Mayr, Franz and Yovine, Sergio and Carrasco, Mat\'ias and Pan, Federico and Vilensky, Federico}, booktitle = {Proceedings of 16th edition of the International Conference on Grammatical Inference}, pages = {250--264}, year = {2023}, editor = {Coste, François and Ouardi, Faissal and Rabusseau, Guillaume}, volume = {217}, series = {Proceedings of Machine Learning Research}, month = {10--13 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v217/mayr23a/mayr23a.pdf}, url = {https://proceedings.mlr.press/v217/mayr23a.html}, abstract = {The paper proposes an approach for probably approximately correct active learning of probabilistic automata (PDFA) from neural language models. It is based on a congruence over strings which is parameterized by an equivalence relation over probability distributions. The learning algorithm is implemented using a tree data structure of arbitrary (possibly unbounded) degree. The implementation is evaluated with several equivalences on LSTM and Transformer-based neural language models from different application domains.} }
Endnote
%0 Conference Paper %T A Congruence-based Approach to Active Automata Learning from Neural Language Models %A Franz Mayr %A Sergio Yovine %A Matías Carrasco %A Federico Pan %A Federico Vilensky %B Proceedings of 16th edition of the International Conference on Grammatical Inference %C Proceedings of Machine Learning Research %D 2023 %E François Coste %E Faissal Ouardi %E Guillaume Rabusseau %F pmlr-v217-mayr23a %I PMLR %P 250--264 %U https://proceedings.mlr.press/v217/mayr23a.html %V 217 %X The paper proposes an approach for probably approximately correct active learning of probabilistic automata (PDFA) from neural language models. It is based on a congruence over strings which is parameterized by an equivalence relation over probability distributions. The learning algorithm is implemented using a tree data structure of arbitrary (possibly unbounded) degree. The implementation is evaluated with several equivalences on LSTM and Transformer-based neural language models from different application domains.
APA
Mayr, F., Yovine, S., Carrasco, M., Pan, F. & Vilensky, F.. (2023). A Congruence-based Approach to Active Automata Learning from Neural Language Models. Proceedings of 16th edition of the International Conference on Grammatical Inference, in Proceedings of Machine Learning Research 217:250-264 Available from https://proceedings.mlr.press/v217/mayr23a.html.

Related Material