Wireless Memory Approximation for Energy-efficient Task-specific IoT Data Retrieval

TL;DR

This paper introduces wireless memory activation and approximation techniques to reduce energy consumption in IoT devices by efficiently managing DRAM-based ML model storage, balancing energy savings with retrieval accuracy.

Contribution

It proposes novel wireless memory management methods specifically designed for energy-efficient IoT data retrieval, addressing DRAM refresh energy waste.

Findings

Significant energy reduction compared to always-on memory schemes.

Maintains retrieval accuracy within specified constraints.

Demonstrates practical feasibility through numerical simulations.

Abstract

The use of Dynamic Random Access Memory (DRAM) for storing Machine Learning (ML) models plays a critical role in accelerating ML inference tasks in the next generation of communication systems. However, periodic refreshment of DRAM results in wasteful energy consumption during standby periods, which is significant for resource-constrained Internet of Things (IoT) devices. To solve this problem, this work advocates two novel approaches: 1) wireless memory activation and 2) wireless memory approximation. These enable the wireless devices to efficiently manage the available memory by considering the timing aspects and relevance of ML model usage; hence, reducing the overall energy consumption. Numerical results show that our proposed scheme can realize smaller energy consumption than the always-on approach while satisfying the retrieval accuracy constraint.