Task-specific node pruning enhances computational efficiency of reservoir computing networks

The relationship between a reservoir network structure and the overall performance of a reservoir computer remains an unanswered question, particularly when searching for an optimal yet small network. In this study, we introduce a systematic, task-specific node-pruning framework that enhances the efficiency while decreasing the size of the reservoir networks. We demonstrate that large networks can be compressed while preserving-or even improving-performance by node removal. Our findings reveal the emergence of optimal subnetwork structures from larger Erdös-Rényi random networks, indicating that efficiency is governed not merely by size but by topological organization. We observe that the pruning process enhances the reservoir's structural efficiency by forming a self-organized asymmetric distribution of input-receiving and readout nodes, with significant changes in key graph-theoretic measures, such as consistent decreases in spectral radius and average degree. Interestingly, the best-performing pruned networks exhibited significantly lower linear memory capacities over the initial networks, which often did not align with the memory demands of the respective tasks. We show that pruning leads to non-uniform network refinements, where specific nodes and connectivity patterns become critical for information flow and task-specific memory retention. This work offers fundamental insights into how structural optimization influences reservoir dynamics, providing a pathway toward designing more efficient, scalable, and interpretable machine learning architectures.