Efficient Approximate Cubing Function Circuits Based on PWL

Cubic operations are frequently used in fields such as digital signal processing and machine learning. In order to improve the efficiency of cubic unit operations, this paper proposes a novel approximation algorithm based on piecewise linear (PWL) segmentation. The approximate cubic operation unit i...

Full description

Saved in:
Bibliographic Details
Main Authors: Zeng, Dongdong, Yuan, Yida, Shi, Weiwei, Liu, Kezhu, Huang, Canji
Format: Conference Proceeding
Language:English
Subjects:
approximation
Approximation algorithms
Conferences
cubic
Digital signal processing
Machine learning
Machine learning algorithms
piecewise linear
Signal processing algorithms
support vector machine
Support vector machines
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Cubic operations are frequently used in fields such as digital signal processing and machine learning. In order to improve the efficiency of cubic unit operations, this paper proposes a novel approximation algorithm based on piecewise linear (PWL) segmentation. The approximate cubic operation unit is applied to Support Vector Machine (SVM). The SVM module is divided into an online part and an offline part. The classification accuracy of the SVM module using the approximate algorithm and the quasi-valued SVM module are 79.17% and 83.3% respectively. This paper designs a 16-bit hardware computing unit through this approximate algorithm. Compared with the previous two algorithms, the area is saved by 76% and 92% respectively, and the delay is also slightly improved. This shows that the proposed approximate algorithm can effectively improve the hardware performance under the premise of ensuring accuracy.
ISSN:2768-6515
DOI:10.1109/ICET55676.2022.9825098