Patent · US Active

Bit width selection for fixed point neural networks

US10262259B2 · kind B2 · utility

1Cited by

5References

30Claims

0Family size

Assignee

QUALCOMM Incorporated · US

Inventors

Dexu Lin · San Diego, US
Venkata Sreekanta Reddy Annapureddy · San Diego, US
David Jonathan Julian · San Diego, US
Casimir Matthew WIERZYNSKI · New York, US

Key dates

Filing date	Nov 9, 2015
Grant date	Apr 16, 2019
Priority date	—
Expiry date	Sep 30, 2037

Classification

Technology area (CPC G)Physics
CPC primaryG06N3/10
WIPO fieldComputer technology
WIPO sectorElectrical engineering

Abstract

A method for selecting bit widths for a fixed point machine learning model includes evaluating a sensitivity of model accuracy to bit widths at each computational stage of the model. The method also includes selecting a bit width for parameters, and/or intermediate calculations in the computational stages of the mode. The bit width for the parameters and the bit width for the intermediate calculations may be different. The selected bit width may be determined based on the sensitivity evaluation.

Source: USPTO / EPO open patent data. Objective bibliographic and citation counts.