TRAM: Training Approximate Multiplier Structures for Low-Power AI Accelerators

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Meng, Chang, Wang, Hanyu, Ye, Yuyang, Yu, Mingfei, Burleson, Wayne, De Micheli, Giovanni
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909028165091328
author Meng, Chang
Wang, Hanyu
Ye, Yuyang
Yu, Mingfei
Burleson, Wayne
De Micheli, Giovanni
author_facet Meng, Chang
Wang, Hanyu
Ye, Yuyang
Yu, Mingfei
Burleson, Wayne
De Micheli, Giovanni
contents Reducing power consumption in AI accelerators is increasingly important. Approximate computing can reduce power consumption while keeping the accuracy loss small. Since multipliers are power-hungry components in AI models, this paper focuses on synthesizing low-power approximate multipliers (AxMs). Unlike prior works that design AxMs separately from AI model training, we present TRAM, which jointly optimizes the AxM structure and AI model parameters to lower power with small accuracy loss. Experiments show that compared to state-of-the-art AxMs, TRAM achieves up to 25.05% AxM power reduction on CNNs with CIFAR-10, and reduces power by up to 27.09% on vision transformers with ImageNet.
format Preprint
id arxiv_https___arxiv_org_abs_2605_08231
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle TRAM: Training Approximate Multiplier Structures for Low-Power AI Accelerators
Meng, Chang
Wang, Hanyu
Ye, Yuyang
Yu, Mingfei
Burleson, Wayne
De Micheli, Giovanni
Machine Learning
Artificial Intelligence
Hardware Architecture
Reducing power consumption in AI accelerators is increasingly important. Approximate computing can reduce power consumption while keeping the accuracy loss small. Since multipliers are power-hungry components in AI models, this paper focuses on synthesizing low-power approximate multipliers (AxMs). Unlike prior works that design AxMs separately from AI model training, we present TRAM, which jointly optimizes the AxM structure and AI model parameters to lower power with small accuracy loss. Experiments show that compared to state-of-the-art AxMs, TRAM achieves up to 25.05% AxM power reduction on CNNs with CIFAR-10, and reduces power by up to 27.09% on vision transformers with ImageNet.
title TRAM: Training Approximate Multiplier Structures for Low-Power AI Accelerators
topic Machine Learning
Artificial Intelligence
Hardware Architecture
url https://arxiv.org/abs/2605.08231