Sample-Efficiency of Kolmogorov-Arnold Networks

2026-10-07

AI Summary

This study addresses the high cost of acquiring real-world samples in reinforcement learning by systematically investigating the sample efficiency of Kolmogorov-Arnold Networks (KAN) in control tasks. Comparative experiments conducted on the Gymnasium benchmarks and the Feynman dataset reveal a novel mechanism through which KAN improves sample utilization, leveraging its inherent parameter efficiency and interpretability. The results demonstrate that, compared to conventional multilayer perceptrons (MLPs), KAN achieves comparable performance using only 60% of the training samples while improving training efficiency by 50%. Furthermore, KAN exhibits strong robustness to reward noise. These findings indicate that KAN provides an effective and efficient solution for reinforcement learning scenarios characterized by costly environmental interactions.

Abstract

Deep reinforcement learning has achieved substantial performance gains over classical control approaches. Yet, a central challenge to learning in real-world applications is acquiring costly samples. Kolmogorov-Arnold Networks are a recently proposed architecture that can learn physical relationships in control problems effectively, with significantly higher parameter efficiency and interpretability when compared to Multi-Layer-Perceptron architectures. In this work, we systematically study sample-efficiency using computational experiments, covering the Feynman dataset and the Gymnasium RL benchmark. The results show that similar performance can be achieved with 40% fewer samples using the Kolmogorov-Arnold architecture, and that relative performance improvements up to 50% occur during the training process. The observed gains are robust to varying levels of noise in rewards. These results highlight the potential of the Kolmogorov-Arnold architectures for more sample-efficient reinforcement learning. Code: https://github.com/DerKevinRiehl/neurips26_kan_training

Problem

• Sample Efficiency
• Deep Reinforcement Learning
• Kolmogorov-Arnold Networks
• Control Problems

Innovation

• Kolmogorov-Arnold Networks
• Sample-Efficiency
• Deep Reinforcement Learning
• Parameter Efficiency
• Robustness