Files in this item



application/pdfECE499-Sp2015-tomei.pdf (383kB)Restricted to U of Illinois
(no description provided)PDF


Title:Approximate Computing with Probabilistic Programs
Author(s):Tomei, Matthew
Subject(s):approximate computing
probabilistic programming
Abstract:Approximate computing involves relaxing program accuracy requirements to improve performance or decrease energy consumption. Since program accuracy measures tend to be non-deterministic due to multiple sources of uncertainty (e.g., inputs), it should be possible to reason about an approximate program as a probabilistic program. In this thesis, we present a framework for reasoning about approximate programs as probabilistic programs. This framework enables solving the problem of finding the optimal approximate program, i.e. the approximate program that minimizes cost while providing statistical guarantees for correctness, in a general setting. Treating approximate programs as probabilistic programs also allows us to apply probabilistic programming tools to approximate computing. One such tool allows us to decrease the time it takes to find an optimal approximate program.
Issue Date:2015-05
Date Available in IDEALS:2015-08-03

This item appears in the following Collection(s)

Item Statistics