Leveraging Language to Learn Program Abstractions and Search Heuristics

Catherine Wong, Kevin M Ellis, Joshua Tenenbaum, Jacob Andreas
Proceedings of the 38th International Conference on Machine Learning, PMLR 139:11193-11204, 2021.

Abstract

Inductive program synthesis, or inferring programs from examples of desired behavior, offers a general paradigm for building interpretable, robust, andgeneralizable machine learning systems. Effective program synthesis depends on two key ingredients: a strong library of functions from which to build programs, and an efficient search strategy for finding programs that solve a given task. We introduce LAPS (Language for Abstraction and Program Search), a technique for using natural language annotations to guide joint learning of libraries and neurally-guided search models for synthesis. When integrated into a state-of-the-art library learning system (DreamCoder), LAPS produces higher-quality libraries and improves search efficiency and generalization on three domains {–} string editing, image composition, and abstract reasoning about scenes {–} even when no natural language hints are available at test time.

Cite this Paper

BibTeX
@InProceedings{pmlr-v139-wong21a, title = {Leveraging Language to Learn Program Abstractions and Search Heuristics}, author = {Wong, Catherine and Ellis, Kevin M and Tenenbaum, Joshua and Andreas, Jacob}, booktitle = {Proceedings of the 38th International Conference on Machine Learning}, pages = {11193--11204}, year = {2021}, editor = {Meila, Marina and Zhang, Tong}, volume = {139}, series = {Proceedings of Machine Learning Research}, month = {18--24 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v139/wong21a/wong21a.pdf}, url = {https://proceedings.mlr.press/v139/wong21a.html}, abstract = {Inductive program synthesis, or inferring programs from examples of desired behavior, offers a general paradigm for building interpretable, robust, andgeneralizable machine learning systems. Effective program synthesis depends on two key ingredients: a strong library of functions from which to build programs, and an efficient search strategy for finding programs that solve a given task. We introduce LAPS (Language for Abstraction and Program Search), a technique for using natural language annotations to guide joint learning of libraries and neurally-guided search models for synthesis. When integrated into a state-of-the-art library learning system (DreamCoder), LAPS produces higher-quality libraries and improves search efficiency and generalization on three domains {–} string editing, image composition, and abstract reasoning about scenes {–} even when no natural language hints are available at test time.} }
Endnote
%0 Conference Paper %T Leveraging Language to Learn Program Abstractions and Search Heuristics %A Catherine Wong %A Kevin M Ellis %A Joshua Tenenbaum %A Jacob Andreas %B Proceedings of the 38th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2021 %E Marina Meila %E Tong Zhang %F pmlr-v139-wong21a %I PMLR %P 11193--11204 %U https://proceedings.mlr.press/v139/wong21a.html %V 139 %X Inductive program synthesis, or inferring programs from examples of desired behavior, offers a general paradigm for building interpretable, robust, andgeneralizable machine learning systems. Effective program synthesis depends on two key ingredients: a strong library of functions from which to build programs, and an efficient search strategy for finding programs that solve a given task. We introduce LAPS (Language for Abstraction and Program Search), a technique for using natural language annotations to guide joint learning of libraries and neurally-guided search models for synthesis. When integrated into a state-of-the-art library learning system (DreamCoder), LAPS produces higher-quality libraries and improves search efficiency and generalization on three domains {–} string editing, image composition, and abstract reasoning about scenes {–} even when no natural language hints are available at test time.
APA
Wong, C., Ellis, K.M., Tenenbaum, J. & Andreas, J.. (2021). Leveraging Language to Learn Program Abstractions and Search Heuristics. Proceedings of the 38th International Conference on Machine Learning, in Proceedings of Machine Learning Research 139:11193-11204 Available from https://proceedings.mlr.press/v139/wong21a.html.

Related Material