Close

Presentation

Statheros: Compiler for Efficient Low-Precision Probabilistic Programming
TimeWednesday, December 8th10:30am - 10:50am PST
Location3014
Event Type
Research Manuscript
Virtual Programs
Presented In-Person
Keywords
Approximate Computing for AI/ML
Topics
Design
DescriptionAs Edge and IoT computing devices process noisy data or make decisions in uncertain environments, they require frameworks for inexpensive, yet accurate probabilistic inference.

We present Statheros, the first compiler for low-level, fixed-point approximation of probabilistic programming. Statheros translates 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.