Data-based analysis of Laplacian Eigenmaps for manifold reduction in supervised Liquid State classifiers

被引：12

作者：

Arena, Paolo ^{[1
]}

Patane, Luca ^{[1
]}

Spinosa, Angelo Giuseppe ^{[1
]}

机构：

[1] Univ Catania, Dipartimento Ingn Elettr Elettron & Informat, Viale Andrea Doria 6, I-95125 Catania, CT, Italy

来源：

INFORMATION SCIENCES | 2019年 / 478卷

关键词：

Reservoir computing; Classification; Laplacian; Manifold reduction; DIMENSIONALITY REDUCTION; DIAGNOSIS;

D O I：

10.1016/j.ins.2018.11.017

中图分类号：

TP [自动化技术、计算机技术];

学科分类号：

0812 ;

摘要：

The manuscript introduces a data-driven technique founded on Laplacian Eigenmaps for manifold reduction in bio-inspired Liquid State classifiers. Starting from a preliminary introduction about the algorithm and the need of using manifold reduction methods for data representation, a statistical analysis of hyperparameters involved in the Laplacian Eigenmaps technique is presented and the effects of quantisation on trained weights is discussed with a view to efficiently implement multiple parallel mappings in the digital domain. (C) 2018 Elsevier Inc. All rights reserved.

引用

页码：28 / 39

页数：12

共 32 条

[1] Neural reuse: A fundamental organizational principle of the brain [J].