IQUIST Master Calendar

Back to Listing

CS Compiler Seminar: Jacob Scott Laurel, "Statheros: A Compiler for Efficient Low-Precision Probabilistic Programming"

Event Type
Seminar/Symposium
Sponsor
Computer Science Department at the University of Illinois
Location
https://illinois.zoom.us/j/89400978467?pwd=NjV3ZFQrQ1JidTNyS0ZNUEVOcEtpUT09
Virtual
wifi event
Date
Oct 25, 2021   4:30 pm  
Speaker
Jacob Scott Laurel, PhD Student, UIUC Compter Science
Contact
Madeleine Garvey
E-Mail
mgarvey@illinois.edu
Views
66
Originating Calendar
Computer Science Speakers Calendar

Speaker: Jacob Scott Laurel

Abstract: As Edge and IoT computing devices process noisy data or make decisions in uncertain environments, they require frameworks for inexpensive, yet accurate probabilistic inference. Probabilistic programming has emerged as a powerful way for developers to write high-level programs, while abstracting away the implementation details of inference. However, the existing algorithms are slow and often assumed to require precise calculations. We present Statheros, the first compiler for low-level, fixed-point approximation of probabilistic programming. Statheros compiles programs to fixed-point inference procedures and is able to determine the optimal fixed-point type to use. We evaluate Statheros on 13 benchmarks and three embedded platforms. The results show that Statheros-generated code is 11.5x (Arduino), 3.8x (PocketBeagle), and 2.2x (Raspberry Pi) faster than single-precision floating-point computation, with minimal accuracy loss.

Bio: Jacob is a 5th year CS PhD student working with Prof. Sasa Misailovic on Probabilistic Programming, Approximate Computing, and more recently static analysis of Neural Networks.

link for robots only