Files in this item



application/pdfUILU-ENG-14-2203.pdf (11MB)Restricted to U of Illinois
(no description provided)PDF


Title:Statistical Error Compensation for Robust Digital Signal Processing and Machine Learning
Author(s):Kim, Eric Park
Machine learning
Statistical computing
Abstract:Machine learning (ML) based inference has recently gained importance as a key kernel in processing massive data in digital signal processing (DSP) systems. Because of the ever-increasing complexity of DSP systems, energy-efficient ML accelerators are critical. Traditionally, energy efficiency was obtained through technology scaling. However, modern nanoscale complementary metal-oxide semiconductor (CMOS) process technologies suffer from reliability problems caused by process, temperature, and voltage variations. As ML applications are inherently probabilistic and robust to errors, statistical error compensation (SEC) techniques can play a significant role in achieving robust and energy-efficient implementation of these important kernels. SEC embraces the statistical nature of errors and utilizes statistical and probabilistic techniques to build robust systems. Energy efficiency is obtained by trading off the enhanced robustness with energy. This dissertation focuses on utilizing statistical approaches via SEC in implementing energy-efficient digital signal processing (DSP) systems with an emphasis on machine learning kernels. Specifically, SEC was applied to a detection-based pseudonoise (PN) code acquisition filter, a communication-centric machine learning kernel, a low-density parity check (LDPC) decoder, and a complex message-passing application, namely a Markov random field (MRF) based stereo image matcher. Results show robust operation up to error rates of 85.83%, while achieving energy savings of 40% to 60%. To further increase energy efficiency and reduce the compensation complexity, higher-level error compensation was explored. Approximate computing (AC) was further combined with SEC, resulting in an additional 5% energy savings, which was enabled through use of algorithms that recognized the statistical nature of the underlying process. Finally, SEC techniques are analyzed to provide insight into the trade-offs in the design of SEC-based systems. Algorithmic noise tolerance is analyzed under a unifying framework based on detection and estimation theory. ANT is shown to approximate the Bayes optimal detector and estimator.
Issue Date:2014-08
Publisher:Coordinated Science Laboratory. University of Illinois at Urbana-Champaign
Series/Report:Coordinated Science Laboratory Report no. UILU-ENG-14-2203
Genre:Technical Report
Sponsor:Microelectronics Advanced Research Corporation (MARCO) / 2013-MA-2385
Date Available in IDEALS:2017-09-06

This item appears in the following Collection(s)

Item Statistics