Reinitializing weights vs units for maintaining plasticity in neural networks

2025-07-31 · J. Fernando Hernandez-Garcia, Shibhansh Dohare, Jun Luo, Rich S. Sutton arxiv

Loss of plasticity is a phenomenon in which a neural network loses its ability to learn when trained for an extended time on non-stationary data. It is a crucial problem to overcome when designing systems that learn continually. An effective technique for preventing loss of plasticity is reinitializing parts of the network. In this paper, we compare two different reinitialization schemes: reinitializing units vs reinitializing weights. We propose a new algorithm, which we name \textit{selective weight reinitialization}, for reinitializing the least useful weights in a network. We compare our algorithm to continual backpropagation and ReDo, two previously proposed algorithms that reinitialize units in the network. Through our experiments in continual supervised learning problems, we identify two settings when reinitializing weights is more effective at maintaining plasticity than reinitializing units: (1) when the network has a small number of units and (2) when the network includes layer normalization. Conversely, reinitializing weights and units are equally effective at maintaining plasticity when the network is of sufficient size and does not include layer normalization. We found that reinitializing weights maintains plasticity in a wider variety of settings than reinitializing units.

📄 PDF Abstract BibTeX arXiv:2508.00212

Code (0)

등록된 구현이 없습니다.

Slow and Steady Wins the Race: Maintaining Plasticity with Hare and Tortoise Networks

2024-06-01 · Hojoon Lee, Hyeonseo Cho, Hyunseung Kim, Donghu Kim 외

This study investigates the loss of generalization ability in neural networks, revisiting warm-starting experiments from Ash & Adams. Our empirical analysis reveals that common methods designed to enhance plasticity by m…

General Knowledge Hippocampus

Spectral Collapse Drives Loss of Plasticity in Deep Continual Learning

2025-09-26 · Arjun Prakash, Naicheng He, Kaicheng Guo, Saket Tiwari 외 arxiv

We investigate why deep neural networks suffer from loss of plasticity in continual learning, and thus fail to learn new tasks without reinitializing parameters. We show that this failure is preceded by Hessian spectral …

Reinforcement Learning Continual Learning

Stagnant Neuron: Towards Understanding the Plasticity Loss in Multi-Agent Reinforcement Learning Value Factorization Methods

2026-06-24 · Zhengzhu Liu, Zeming Gao, Haoyuan Qin, Jiawei Hu 외 arxiv

Multi-Agent Reinforcement Learning (MARL) value factorization methods can suffer from a loss of plasticity, gradually failing to adapt when transferring to new task instances. We trace this issue to stagnant neurons, uni…

Multi-agent Reinforcement Learning

Normalisation of Weights and Firing Rates in Spiking Neural Networks with Spike-Timing-Dependent Plasticity

2019-09-30 · Katarzyna Kozdon, Peter Bentley

Maintaining the ability to fire sparsely is crucial for information encoding in neural networks. Additionally, spiking homeostasis is vital for spiking neural networks with changing numbers of weights and neurons. We dis…

Clustering Time Series Time Series Analysis

CBPNet: A Continual Backpropagation Prompt Network for Alleviating Plasticity Loss on Edge Devices

2025-09-19 · Runjie Shao, Boyu Diao, Zijia An, Ruiqi Liu 외 arxiv

To meet the demands of applications like robotics and autonomous driving that require real-time responses to dynamic environments, efficient continual learning methods suitable for edge devices have attracted increasing …

Continual Learning Autonomous Driving