Have a UROP opening you would like to submit?
Please fill out the form.
Answering Why? Explanation systhemsis through causality, probabilsitic inference and program induction. adsa
6: Electrical Engineering and Computer Science
Armando Solar Lezama
Constructing and communicating explanations is a fundamental and routine human capacity. Despite advances in machine learning, no existing system comes close to approximating these capabilities. A system that constructs explanations must address three problems simultaneously: The objective of this project is to construct a general system that can answer "why?" questions, that is, construct explanations of observations. In particular, the first domain will be physics-based games. Given an event (e.g. the player died), the system will construct explanations that resemble what a human would give (e.g. the player accelerated too fast around the first corner). - Probabilistic inference to infer latent world parameters (e.g. the 3d shape and object positions) from observed data (e.g. 2D pixel data) - Program Synthesis - Causal Inference This project is led by Zenna Tavares (firstname.lastname@example.org) under supervision of Armando Solar Lezama.
This is an exciting project. We expect candidates to have expertise in at least some of the following fields: - Physics simulation - Graphics/rendering - Probabilistic inference - Causal inference - Programming languages and program analysis - Program Synthesis/induction - Programming languages: Julia, Python, C++