My dissertation, Look me up at DBLP for another view on my publications.
Some of the links below are to the ACM database (so they can affect the ACM’s popularity statistics).
Floating-point numbers, for example, approximate real-number arithmetic to save space and time over arbitrary-precision numerical representation.
This document focuses on the study of approximate computing and system-level techniques that apply this theory to create new trade-offs.
The general idea is to spend less energy on retaining or accessing data; in return, there is a small probability that bits will flip in the memory.
Drg Assignment - Arjang Hassibi Thesis
SRAM structures spend significant static power on retaining data, so they represent another opportunity for fidelity trade-offs [@hybrid-sram; @sramerrors; @partially-forgetful].
This section deals with hardware-oriented approximation techniques.
We categorize the techniques according to the hardware component they affect.
It's a living document meant to exhaustively catalog everything we know about approximation along with the earlier work that influenced it.
It's also a collaborative, open-source project: to contribute, see its home on Git Hub.