arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2509.22011stat.MLcs.LGmath.STstat.TH
收藏

回声状态网络的随机矩阵视角:从精确偏差—方差刻画到最优正则化

A Random Matrix Perspective of Echo State Networks: From Precise Bias--Variance Characterization to Optimal Regularization

Huawei Noah’s Ark Lab, Huawei Technologies, Paris, France(华为诺亚实验室,华为技术,法国巴黎);Chinese university of Hongkong, Shenzhen, China(香港中文大学(深圳));Huazhong University of Science Technology (HUST), Wuhan, China(华中科技大学)

Yessin Moakher, Malik Tiomoko, Cosme Louart, Zhenyu Liao

AI总结:

本文利用随机矩阵理论刻画线性教师—学生设定下 ESN 的渐近偏差、方差与 MSE,揭示其无双重下降且小样本、短记忆时更优,并给出最优岭正则化公式及通用数值算法。

摘要:

我们在教师—学生设定下,对具有 oracle 权重的线性教师的 Echo State Networks(ESNs,回声状态网络)进行了严格的渐近分析。利用随机矩阵理论,我们推导了渐近偏差、方差和均方误差(MSE)的闭式表达式,将其表示为输入统计量、oracle 向量以及岭正则化参数的函数。分析揭示了与经典岭回归的两个关键差异:(i)ESN 不表现出双下降;(ii)当训练样本数量和教师记忆长度均有限时,ESN 可获得更低的 MSE。我们进一步给出了单位输入协方差情形下最优正则化的显式公式,并提出了一种在一般情形下计算该最优值的高效数值方案。这些结果共同为调节 ESN 提供了可解释的理论和实用准则,有助于调和近期的经验观察与可证明的性能保证。

英文摘要:

We present a rigorous asymptotic analysis of Echo State Networks (ESNs) in a teacher student setting with a linear teacher with oracle weights. Leveraging random matrix theory, we derive closed form expressions for the asymptotic bias, variance, and mean-squared error (MSE) as functions of the input statistics, the oracle vector, and the ridge regularization parameter. The analysis reveals two key departures from classical ridge regression: (i) ESNs do not exhibit double descent, and (ii) ESNs attain lower MSE when both the number of training samples and the teacher memory length are limited. We further provide an explicit formula for the optimal regularization in the identity input covariance case, and propose an efficient numerical scheme to compute the optimum in the general case. Together, these results offer interpretable theory and practical guidelines for tuning ESNs, helping reconcile recent empirical observations with provable performance guarantees

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑