CDRL: A Reinforcement Learning Framework Inspired by Cerebellar Circuits and Dendritic Computational Strategies

2026-02-17

AI Summary

Reinforcement learning often struggles with low sample efficiency, poor robustness, and weak generalization in high-dimensional, partially observable, and noisy environments. This work proposes a novel reinforcement learning architecture inspired by cerebellar circuitry, systematically incorporating cerebellar structural priors and dendritic computation mechanisms as inductive biases for the first time. By leveraging large-scale expansion, sparse connectivity, sparse activation, and nonlinear dendritic modulation, the resulting agent achieves both biological plausibility and high learning efficiency. Empirical evaluations demonstrate that the proposed method significantly outperforms existing approaches across multiple high-dimensional, noisy benchmark tasks, achieving consistent improvements in sample efficiency, robustness, and generalization—while maintaining superior performance even under strict parameter constraints.

Technology Category

• Machine Learning: Reinforcement Learning
• Cognitive Modeling & Cognitive Systems: Agent Architectures
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Incentives in network design for Web infrastructures and ecosystems
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Reinforcement learning (RL) has achieved notable performance in high-dimensional sequential decision-making tasks, yet remains limited by low sample efficiency, sensitivity to noise, and weak generalization under partial observability. Most existing approaches address these issues primarily through optimization strategies, while the role of architectural priors in shaping representation learning and decision dynamics is less explored. Inspired by structural principles of the cerebellum, we propose a biologically grounded RL architecture that incorporate large expansion, sparse connectivity, sparse activation, and dendritic-level modulation. Experiments on noisy, high-dimensional RL benchmarks show that both the cerebellar architecture and dendritic modulation consistently improve sample efficiency, robustness, and generalization compared to conventional designs. Sensitivity analysis of architectural parameters suggests that cerebellum-inspired structures can offer optimized performance for RL with constrained model parameters. Overall, our work underscores the value of cerebellar structural priors as effective inductive biases for RL.

Problem

• sample efficiency
• noise sensitivity
• generalization
• partial observability
• reinforcement learning

Innovation

• cerebellar-inspired architecture
• dendritic modulation
• sparse connectivity
• sample efficiency
• inductive bias