Event
Ph.D. Dissertation Defense: Mingze Gao
Thursday, April 5, 2018
11:30 a.m.-2:00 p.m.
AVW 2168
Maria Hoo
301 405 3681
mch@umd.edu
ANNOUNCEMENT: Ph.D. Dissertation Defense
Name: Mingze Gao
Committee:
Professor Gang Qu, Chair/Advisor
Professor Shuvra S. Bhattacharyya
Professor Manoj Franklin
Professor Donald Yeung
Professor Yu Chen, Dean's Representative
Date/time: Thursday, April 5th, 2018 at 11:30 am - 2:00 pm.
Place: AVW 2168
Title: Approximate Computing Techniques for Low Power and Energy Efficiency
Abstract:
Approximate computing is an emerging computation paradigm in the era of the Internet of things, big data and AI. It takes advantages of the error-tolerable feature of many applications, such as machine learning and image/signal processing, to reduce the resources consumption and delivers a certain level of computation quality.
In this dissertation, we focus on utilizing the characteristics of different data formats for arithmetic and numerical approximation. Currently, there are mainly three data formats that are widely used in the modern computing system, the integer and fixed point format, and the floating point format. For the integer and fixed point format, we propose a dynamic solution and a static solution. In the dynamic way, we propose an Approximate Integer Format (AIF) and corresponding computing mechanism. In the static way, we analyze the data range to statistically select the MSBs during computations. For the floating point system, we propose a runtime estimation technique by converting data into the logarithmic domain to assess the intermediate results. Then we replace the non-critical computations with the estimated value.
Besides the low power and energy efficiency concern, we propose an approximate floating point based security primitive that enables us to embed information during the process of approximate computing. The hidden information can be generated, embedded, and retrieved for several security applications.