Sample-Efficiency of Kolmogorov-Arnold Networks

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.