Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Sparse Optimization for Green Edge AI Inference

Xiangyu Yang, Sheng Hua, Yuanming Shi, Hao Wang, Jun Zhang, Khaled B. Letaief

TL;DR

This paper presents a joint inference task selection and downlink beamforming strategy to achieve energy-efficient edge AI inference through minimizing the overall power consumption consisting of both computation and transmission power consumption, yielding a mixed combinatorial optimization problem.

Abstract

With the rapid upsurge of deep learning tasks at the network edge, effective edge artificial intelligence (AI) inference becomes critical to provide low-latency intelligent services for mobile users via leveraging the edge computing capability. In such scenarios, energy efficiency becomes a primary concern. In this paper, we present a joint inference task selection and downlink beamforming strategy to achieve energy-efficient edge AI inference through minimizing the overall power consumption consisting of both computation and transmission power consumption, yielding a mixed combinatorial optimization problem. By exploiting the inherent connections between the set of task selection and group sparsity structural transmit beamforming vector, we reformulate the optimization as a group sparse beamforming problem. To solve this challenging problem, we propose a log-sum function based three-stage approach. By adopting the log-sum function to enhance the group sparsity, a proximal iteratively reweighted algorithm is developed. Furthermore, we establish the global convergence analysis and provide the ergodic worst-case convergence rate for this algorithm. Simulation results will demonstrate the effectiveness of the proposed approach for improving energy efficiency in edge AI inference systems.

Sparse Optimization for Green Edge AI Inference

TL;DR

This paper presents a joint inference task selection and downlink beamforming strategy to achieve energy-efficient edge AI inference through minimizing the overall power consumption consisting of both computation and transmission power consumption, yielding a mixed combinatorial optimization problem.

Abstract

With the rapid upsurge of deep learning tasks at the network edge, effective edge artificial intelligence (AI) inference becomes critical to provide low-latency intelligent services for mobile users via leveraging the edge computing capability. In such scenarios, energy efficiency becomes a primary concern. In this paper, we present a joint inference task selection and downlink beamforming strategy to achieve energy-efficient edge AI inference through minimizing the overall power consumption consisting of both computation and transmission power consumption, yielding a mixed combinatorial optimization problem. By exploiting the inherent connections between the set of task selection and group sparsity structural transmit beamforming vector, we reformulate the optimization as a group sparse beamforming problem. To solve this challenging problem, we propose a log-sum function based three-stage approach. By adopting the log-sum function to enhance the group sparsity, a proximal iteratively reweighted algorithm is developed. Furthermore, we establish the global convergence analysis and provide the ergodic worst-case convergence rate for this algorithm. Simulation results will demonstrate the effectiveness of the proposed approach for improving energy efficiency in edge AI inference systems.

Paper Structure

This paper contains 26 sections, 8 theorems, 62 equations, 4 figures, 2 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Related Works
  3. Organization
  4. Notation
  5. System Model and Problem Formulation
  6. System Model
  7. Downlink Transmission Model
  8. Power Consumption Model
  9. Problem Formulation and Analysis
  10. A Group Sparse Beamforming Representation Framework
  11. A Los-sum Function Based Three-stage Group Sparse Beamforming Framework
  12. Log-sum Function for Enhancing Group Sparsity
  13. A Log-sum Function Based Three-stage Group Sparse Beamforming Approach
  14. Global Convergence Analysis
  15. First-order Necessary Optimality Condition
  16. ...and 11 more sections

Key Result

Proposition 1

Let $X$ be a closed and convex set. Then the following properties on Fréchet subdifferentials holds true.

Figures (4)

  • Figure 1: System model illustration of the edge AI inference for intelligent tasks. This paper considers the scenario that each neighboring edge BS has collected the raw input data $\{\bm{d}_{k}\}$ from mobile users.
  • Figure 2: The estimated energy consumption breakdown EstimationWebsite of the GoogLeNet v1 to perform image classification tasks on the Eyeriss chip chen2016eyeriss.
  • Figure 3: Convergence of the proximal iteratively reweighted algorithm for log-sum minimization problem.
  • Figure 4: Average total network power consumption comparison for three different approaches in edge AI inference system.

Theorems & Definitions (9)

  • Definition 1: Fréchet subdifferential kruger2003frechet
  • Proposition 1
  • Theorem 1: Fermat's rule
  • Proposition 2
  • Lemma 1
  • Lemma 2
  • Theorem 2
  • Lemma 3
  • Theorem 3