发表机构
Sungkyunkwan University（成均馆大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
该研究针对回声状态网络储备池过参数化问题，提出动力学模式剪枝方法，通过轨迹平均雅可比格拉姆矩阵排序神经元，实验显示其可提升或保持预测精度并减少冗余组件。
AI 中文摘要
回声状态网络（Echo State Networks，ESN）是一种用于时间序列预测的高效框架，但其随机初始化的储备池通常存在过参数化和动态冗余的问题。现有剪枝方法大多依赖静态连接性或激活统计，可能忽略那些塑造输入驱动状态转换的神经元。我们提出动力学模式剪枝（Dynamical Mode Pruning，DMP）这一储备池剪枝方法，该方法根据神经元对从轨迹平均雅可比格拉姆矩阵中获得的主导转换模式的贡献对神经元进行排序。DMP会移除低影响力单元，仅重新训练读出层。在混沌和真实世界时间序列基准上的实验表明，DMP在降低冗余储备池组件的同时，可提升或保持预测精度。我们的结果表明，动力学影响力是储备池优化的有用准则，其价值超出了单纯的静态结构重要性。
英文摘要
Echo State Networks (ESNs) offer an efficient framework for temporal prediction, but their randomly initialized reservoirs are often over-parameterized and dynamically redundant. Existing pruning methods largely rely on static connectivity or activation statistics, which may overlook neurons that shape input-driven state transitions. We propose Dynamical Mode Pruning (DMP), a reservoir pruning method that ranks neurons by their contribution to dominant transition modes obtained from a trajectory-averaged Jacobian Gramian. DMP removes low-impact units and retrains only the readout. Experiments on chaotic and real-world time-series benchmarks show that DMP improves or preserves forecasting accuracy while reducing redundant reservoir components. Our results suggest that dynamical influence is a useful criterion for reservoir refinement beyond static structural importance alone.
Comments18 pages, 6 figures
