Ketaki Joshi

Ketaki Joshi

Ph.D Candidate, Computer Science

Distill: Domain-Specific Compilation for Cognitive Models

Computational models of cognition enable a better understanding of the human brain and behavior, psychiatric and neurological illnesses, and clinical interventions to treat illnesses. They offer a path towards human-like artificial intelligence. Cognitive models are laborious to develop, requiring the composition of many types of computational tasks. They suffer from poor performance as they are generally designed using high-level languages like Python. In this work, we present Distill, a domain-specific compilation tool to accelerate cognitive models while continuing to offer cognitive scientists the ability to develop their models in flexible high-level languages.
Distill uses domain-specific knowledge to compile Python-based cognitive models into LLVM IR, carefully stripping away features like dynamic typing and memory management that add performance overheads without being necessary for the underlying computation of the models. The net effect is a performance improvement in model execution by Distil over state-of-the-art techniques using Pyston and PyPy.
Distill also repurposes classical compiler data flow analyses to reveal properties of data flow in cognitive models that are useful to cognitive scientists such as detecting computational clones to enable reusability of models. Distill identifies computational clones in different cognitive models to enable reuse or substitution of complex models by simpler ones. Distill is publicly available, integrated with the PsyNeuLink cognitive modeling environment, and is already being used by researchers in the brain sciences.
Paper:
Ján Veselý, Raghavendra Pradyumna Pothukuchi, Ketaki Joshi, Samyak Gupta, Jonathan D. Cohen, and Abhishek Bhattacharjee. 2022. Distill: domain-specific compilation for cognitive models. In Proceedings of the 20th IEEE/ACM International Symposium on Code Generation and Optimization (CGO '22). IEEE Press, 301–312. https://doi.org/10.1109/CGO53902.2022.9741278