Thomas Leech

Moscow Institute of Thermal Technology, MIT Lincoln Laboratory

Papers

Total Citations

H-Index

About

Thomas Leech’s research lies at the intersection of machine learning, robotics, and formal logic, with a focus on making autonomous systems both interpretable and controllable. His major contributions center on developing methods that learn imitative policies from expert demonstrations while preserving transparency—a critical challenge in human-robot interaction. In his most cited work, “Deep Bayesian Nonparametric Learning of Rules and Plans from Demonstrations with a Learned Automaton Prior” (2020, 6 citations), Leech introduced a novel approach that models high-level action sequences as automata with connections to formal logic. This allows learned policies to be not only interpretable (users can understand the decision-making structure) but also manipulable (users can modify or correct behavior). His subsequent work, “Learning and planning with logical automata” (2021, 2 citations), further refines these ideas, while “Explainable machine learning for task planning in robotics” (2019, 2 citations) establishes the foundational framework. Though early in his career, Leech’s work is notable for bridging Bayesian nonparametrics with symbolic reasoning, offering a path toward robots that can explain their plans and adapt to human feedback—a key step for safe, trustworthy autonomy.