Hybrid Compute-in-Memory Architecture for Edge AI: Navigating the Precision-Power-Latency Through Optimization Techniques
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
The demand for energy-efficient edge AI hardware has shown the limitations of conventional digital architectures in achieving the desired performance per watt under tight power constraints. This paper examines the recent progress in hybrid analog-digital AI accelerator architectures developed to improve energy efficiency by retaining computational accuracy and scalability. The main advancement are analog compute-in-memory (AIMC) approaches, which minimize data movement and energy consumption by executing multiple accumulate operations directly within memory arrays. This review also discusses techniques such as DVFS and AI-driven runtime power management, which support adaptive workload-aware energy optimization across diverse application contexts. Additionally, the integration of ultralow-bit quantized neural networks, including 4-bit and binary models, improves computational throughput and energy efficiency in resource-constrained environments such as IoT nodes, wearables, and autonomous sensors. Comparative analyses shows that analog-digital codesigns continuously deliver significant energy and latency improvements over traditional digital accelerators.
Fields of Study & Subjects
Advanced Memory and Neural Computing
Low-power high-performance VLSI design
Advanced Neural Network Applications
Publication details
- Journal / source
- International Conference on Intelligent Systems for Communication, IoT and Security (ICISCoIS)
- Pages
- 1–7
- Publication date
- 8 January 2026
- Language
- EN