Learning Deterministic Finite Automata from Infinite Alphabets

Gaetano Pellegrino, Christian Hammerschmidt, Qin Lin, Sicco Verwer
; Proceedings of The 13th International Conference on Grammatical Inference, PMLR 57:120-131, 2017.

Abstract

We proposes an algorithm to learn automata infinite alphabets, or at least too large to enumerate. We apply it to define a generic model intended for regression, with transitions constrained by intervals over the alphabet. The algorithm is based on the Red & Blue framework for learning from an input sample. We show two small case studies where the alphabets are respectively the natural and real numbers, and show how nice properties of automata models like interpretability and graphical representation transfer to regression where typical models are hard to interpret.

Cite this Paper


BibTeX
@InProceedings{pmlr-v57-pellegrino16, title = {Learning Deterministic Finite Automata from Infinite Alphabets}, author = {Gaetano Pellegrino and Christian Hammerschmidt and Qin Lin and Sicco Verwer}, booktitle = {Proceedings of The 13th International Conference on Grammatical Inference}, pages = {120--131}, year = {2017}, editor = {Sicco Verwer and Menno van Zaanen and Rick Smetsers}, volume = {57}, series = {Proceedings of Machine Learning Research}, address = {Delft, The Netherlands}, month = {05--07 Oct}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v57/pellegrino16.pdf}, url = {http://proceedings.mlr.press/v57/pellegrino16.html}, abstract = {We proposes an algorithm to learn automata infinite alphabets, or at least too large to enumerate. We apply it to define a generic model intended for regression, with transitions constrained by intervals over the alphabet. The algorithm is based on the Red & Blue framework for learning from an input sample. We show two small case studies where the alphabets are respectively the natural and real numbers, and show how nice properties of automata models like interpretability and graphical representation transfer to regression where typical models are hard to interpret.} }
Endnote
%0 Conference Paper %T Learning Deterministic Finite Automata from Infinite Alphabets %A Gaetano Pellegrino %A Christian Hammerschmidt %A Qin Lin %A Sicco Verwer %B Proceedings of The 13th International Conference on Grammatical Inference %C Proceedings of Machine Learning Research %D 2017 %E Sicco Verwer %E Menno van Zaanen %E Rick Smetsers %F pmlr-v57-pellegrino16 %I PMLR %J Proceedings of Machine Learning Research %P 120--131 %U http://proceedings.mlr.press %V 57 %W PMLR %X We proposes an algorithm to learn automata infinite alphabets, or at least too large to enumerate. We apply it to define a generic model intended for regression, with transitions constrained by intervals over the alphabet. The algorithm is based on the Red & Blue framework for learning from an input sample. We show two small case studies where the alphabets are respectively the natural and real numbers, and show how nice properties of automata models like interpretability and graphical representation transfer to regression where typical models are hard to interpret.
RIS
TY - CPAPER TI - Learning Deterministic Finite Automata from Infinite Alphabets AU - Gaetano Pellegrino AU - Christian Hammerschmidt AU - Qin Lin AU - Sicco Verwer BT - Proceedings of The 13th International Conference on Grammatical Inference PY - 2017/01/16 DA - 2017/01/16 ED - Sicco Verwer ED - Menno van Zaanen ED - Rick Smetsers ID - pmlr-v57-pellegrino16 PB - PMLR SP - 120 DP - PMLR EP - 131 L1 - http://proceedings.mlr.press/v57/pellegrino16.pdf UR - http://proceedings.mlr.press/v57/pellegrino16.html AB - We proposes an algorithm to learn automata infinite alphabets, or at least too large to enumerate. We apply it to define a generic model intended for regression, with transitions constrained by intervals over the alphabet. The algorithm is based on the Red & Blue framework for learning from an input sample. We show two small case studies where the alphabets are respectively the natural and real numbers, and show how nice properties of automata models like interpretability and graphical representation transfer to regression where typical models are hard to interpret. ER -
APA
Pellegrino, G., Hammerschmidt, C., Lin, Q. & Verwer, S.. (2017). Learning Deterministic Finite Automata from Infinite Alphabets. Proceedings of The 13th International Conference on Grammatical Inference, in PMLR 57:120-131

Related Material