University of Twente Student Theses

Login

Approximate Least Squares Accelerator

Krapukhin, Alexander (2019) Approximate Least Squares Accelerator.

[img] PDF
6MB
Abstract:Approximate computing allows to reduce power, area or increase the speed of a circuit by simplifying its logic. This simplification introduces errors in computations. Some applications can tolerate a certain degree of inaccuracy and the correct computations are not necessary to produce acceptable results. In this work, approximate computing methods are applied to the least squares problem. In terms of hardware required, the least squares computation consists of a multiplier-accumulator (MAC), squarer-accumulator (SAC) and a divider. As the SAC is a special simplified version of the MAC, and the division is typically done only once at the end of the computation, the MAC unit approximation is the most important to analyze. Several approximate techniques are applied to the MAC and their effectiveness is compared. The results of this MAC analysis are used to approximate the more complex least squares unit. As a case study where the approximation of the least squares is applied, radio-astronomy calibration is used. The calibration is performed by iteratively solving the least squares problem to estimate complex antenna gains. This algorithm does not require the same precision during its computation, which allows to map some number of initial iterations to a lower-precision (or approximate) hardware in order to decrease the energy consumption.
Item Type:Essay (Master)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:53 electrotechnology, 54 computer science
Programme:Embedded Systems MSc (60331)
Link to this item:http://purl.utwente.nl/essays/77225
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page