LEAP: Low-Energy AI For Edge Learning and Processing
August 31-September 3 • Istanbul, Türkiye
Special Session at IEEE MLSP 2025 in Istanbul, Türkiye
ABOUT
The increasing pervasiveness of edge devices (as the Internet of Things) for AI-based services (e.g., healthcare, structural health monitoring, environmental monitoring, wireless communication) and the high availability, velocity, and volatility of data generated and collected at the edge of the Internet are pushing toward a paradigm shift in the design of AI-based systems. Machine Learning (ML) and data processing are moving from powerful and remote data centers to distributed systems at the edge, working in proximity to where data are physically generated and/or collected. Differently from cloud data centers, distributed edge solutions are often deployed in resource-constrained devices, which limits the direct implementation of current state-of-the-art, energy-intensive ML models.
CALL FOR PAPERS
This special session at IEEE MLSP 2025 seeks to foster the development of innovative ML solutions that prioritize resource efficiency, enabling high-performance computation on edge devices while ensuring scalability and sustainability. Aligned with this year’s conference theme, this session emphasizes practical contributions that enable ML solutions in real-world, resource-limited environments. We welcome contributions that enable efficient, scalable and sustainable on-device training/inference of traditional and new ML approaches.
Furthermore, this special session promotes the development of sustainable distributed approaches to address the escalating energy demands of centralized ML solutions. By encouraging submissions in energy-aware algorithmic approaches, in-memory computing hardware, communication and brain-inspired models, the session seeks to highlight energy-efficient solutions that bridge the gap between state-of-the-art ML techniques and their deployment in resource constrained settings.
The main topics of interests include (but are not limited to):
- Lightweight and energy-efficient machine learning models for signal processing.
- Efficient ML solutions for on-device training/inference.
- Memristive hardware for energy-efficient computation.
- Adaptive and scalable learning algorithms for dynamic edge environments.
- Communication-efficient protocols for decentralized learning.
- Low-energy ML solutions.
- Decentralized Learning.
Important Dates*
Paper Submission: May 10, 2025
Author Notification: June 24, 2025
Camera-ready Due: July 15, 2025
Workshop Date: TBD
* All deadlines are in AoE time, 23:59.(check it here).
SUBMISSION GUIDELINES
We follow the same paper submission guidelines as the main track of the conference. Papers must not be longer than 6 pages, including all text, figures, and references in accordance with IEEE Signal Processing Society policy and procedures. Please note that MLSP uses a double-blind review process. Thus, it is important that all potentially identifying information (such as author names and affiliations) is removed from the initial submission pdf file. Papers that do not follow this requirement may suffer automatic desk rejection.
To prepare your manuscript, use the formatting guidelines provided in the: IEEE MLPS Submission Guidelines.
Submission link: TBD
REGISTRATION
TBD
