Language-informed Sequential Decision-making

Please enable javascript in your browser.

Hashemi Dijujin, Negin | 2022

74 Viewed

Type of Document: M.Sc. Thesis
Language: Farsi
Document No: 55649 (19)
University: Sharif University of Technology
Department: Computer Engineering
Advisor(s): Soleymani Baghshah, Mahdieh
Abstract:
Sample efficiency and systematic generalization are two long-standing challenges in sequential decision-making problems, especially, in reinforcement learning settings. It is hypothesized that involving natural language in conjunction with other observation modalities in decision-making environments can improve generalization due to its compositional and open-ended nature, and sample efficiency due to the concise information summarized in relatively short linguistic units. By exploiting this information and the compositional structure of the language, one can achieve an abstract and factored understanding of the environment and the task at hand. To do so, it is necessary to find the proper grounding between meaningful data components from different modalities present in the input; e.g. visual and linguistic. In this project, we attempt to examine architecture-level inductive biases that can help improve language-informed reinforcement learning criteria, based on Decision Transformers and Neural Production Systems. Our experiments in BabyAI environments achieve higher sample efficiency and compositional generalization compared to baseline models.
Keywords:
Reinforcement Learning ; Transformers ; Sequential Decision Making ; Language Informed Agents ; Neural Production Systems