Electrical Engineering and Systems Science > Systems and Control
[Submitted on 13 Mar 2024 (v1), last revised 26 Apr 2024 (this version, v2)]
Title:Probabilistic Metaplasticity for Continual Learning with Memristors
View PDF HTML (experimental)Abstract:Edge devices operating in dynamic environments critically need the ability to continually learn without catastrophic forgetting. The strict resource constraints in these devices pose a major challenge to achieve this, as continual learning entails memory and computational overhead. Crossbar architectures using memristor devices offer energy efficiency through compute-in-memory and hold promise to address this issue. However, memristors often exhibit low precision and high variability in conductance modulation, rendering them unsuitable for continual learning solutions that require precise modulation of weight magnitude for consolidation. Current approaches fall short to address this challenge directly and rely on auxiliary high-precision memory, leading to frequent memory access, high memory overhead, and energy dissipation. In this research, we propose probabilistic metaplasticity, which consolidates weights by modulating their update probability rather than magnitude. The proposed mechanism eliminates high-precision modification to weight magnitudes and, consequently, the need for auxiliary high-precision memory. We demonstrate the efficacy of the proposed mechanism by integrating probabilistic metaplasticity into a spiking network trained on an error threshold with low-precision memristor weights. Evaluations of two continual learning benchmarks show that probabilistic metaplasticity achieves state-of-the-art performance while consuming ~ 67% lower memory for additional parameters and up to two orders of magnitude lower energy during parameter updates compared to an auxiliary memory-based solution. The proposed model shows potential for energy-efficient continual learning with low-precision emerging devices.
Submission history
From: Fatima Tuz Zohora [view email][v1] Wed, 13 Mar 2024 17:18:39 UTC (1,704 KB)
[v2] Fri, 26 Apr 2024 17:34:32 UTC (1,724 KB)
Current browse context:
eess.SY
References & Citations
Bibliographic and Citation Tools
Bibliographic Explorer (What is the Explorer?)
Litmaps (What is Litmaps?)
scite Smart Citations (What are Smart Citations?)
Code, Data and Media Associated with this Article
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub (What is DagsHub?)
Gotit.pub (What is GotitPub?)
Papers with Code (What is Papers with Code?)
ScienceCast (What is ScienceCast?)
Demos
Recommenders and Search Tools
Influence Flower (What are Influence Flowers?)
Connected Papers (What is Connected Papers?)
CORE Recommender (What is CORE?)
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.