Special Issue on Approximate Computing: the need for efficient and sustainable computing Submission Date: 2025-01-31 Motivation and Scope

Today, computing systems face unprecedented computational demands. They serve as bridges between the digital and physical world, processing vast data from diverse sources. Our digital world is constantly producing an immense volume of data. According to recent estimates, millions of terabytes of data are created each day. To handle this immense volume of data, increasingly sophisticated and resource-constrained devices are deployed at the edge, where energy and power efficiency takes center stage. Additionally, the growth of modern AI models, particularly neural networks, has led to boundless computational and power demands. For example, GPT-3 featuring 175 billion parameters, BERT large equipped with 340 million parameters, require high energy costs for training and inferencing data. Approximate Computing (AxC) provides a promising solution: by intentionally allowing slight inaccuracies in computations, AxC significantly reduces overhead (including energy, area, and latency) while preserving practical accuracy levels. These paradigms find applications across several domains. With the intent of navigating the intricate balance between accuracy, reliability, and energy efficiency, exploring Approximate Computing (AxC) techniques becomes crucial.

The proposed Special Issue (SI) investigates the intersection of energy-efficient computing and accuracy of state-of-the-art workloads, shedding light on innovative approaches and practical implementations.

The potential areas of interest for the proposed SI include, but are not limited to, the following topics:

Approximation for Deep Learning applications, including Large Language Models (LLMs)

Approximation techniques for emerging processor and memory technologies

Approximation-induced error modeling and propagation

Approximation in edge computing applications

Approximation in HPC and embedded systems

Approximation in Foundation Models

Approximation in reconfigurable computing

Architectural support for approximation

Cross-layer approximate computing

Hardware/software co-design of approximate systems

Dependability of approximate circuits and systems

Design automation of approximate architectures

Design of approximate reconfigurable architectures

Error resilient Near-Threshold Computing

Methods for monitoring and controlling approximation quality

Modeling, specification, and verification of approximate circuits, and systems

Safety and reliability applications of approximate computing

Security in the context of approximation

Software-based fault-tolerant technique for approximate computing

Test and fault tolerance of approximate systems

Guest Editors

Annachiara Ruospo

Politecnico di Torino, Italy

annachiara.ruospo@polito.it

Salvatore Barone

University of Naples Federico II, Italy

salvatore.barone@unina.it

Jorge Castro-Godinez

School of Electronics Engineering

Instituto Tecnologico de Costa Rica, Costa Rica

jocastro@itcr.ac.cr

Important Dates

Submission portal opens: November 1st, 2024

Deadline for paper submission: January 31st, 2025

Latest acceptance deadline for all papers: May 31st, 2025