Automated Circuit Approximation Method Driven by Data Distribution

conference paper

English

We propose an application-tailored data-driven fully automated method for functional approximation of combinational circuits. We demonstrate how an application-level error metric such as the classification accuracy can be translated to a component-level error metric needed for an efficient and fast search in the space of approximate low-level components that are used in the application. This is possible by employing a weighted mean error distance (WMED) metric for steering the circuit approximation process which is conducted by means of genetic programming. WMED introduces a set of weights (calculated from the data distribution measured on a selected signal in a given application) determining the importance of each input vector for the approximation process. The method is evaluated using synthetic benchmarks and application-specific approximate MAC (multiply-and-accumulate) units that are designed to provide the best trade-offs between the classification accuracy and power consumption of two image classifiers based on neural networks.

digital circuit, approximate circuit, functional approximation, neural network

VAŠÍČEK, Z.; MRÁZEK, V.; SEKANINA, L.

26. 3. 2019

European Design and Automation Association

Florence

978-3-9819263-2-3

Design, Automation and Test in Europe Conference

96

101

6

https://www.fit.vut.cz/research/publication/11821/

0099.pdf
date19_final.pdf