Neural networks with a continuous squashing function in the output are universal approximators

被引：51
|
作者：
Castro, JL [1 ]
Mantes, CJ [1 ]
Benítez, JM [1 ]
机构：
[1] Univ Granada, Dept Comp Sci & AI, ETSI Informat, E-18071 Granada, Spain
来源：
NEURAL NETWORKS | 2000年 / 13卷 / 06期
关键词：
feedforward networks; universal approximation; squashing functions; continuous functions;
D O I：
10.1016/S0893-6080(00)00031-9
中图分类号：
TP18 [人工智能理论];
学科分类号：
081104 ; 0812 ; 0835 ; 1405 ;
摘要：
In 1989 Hornik as well as Funahashi established that multilayer feedforward networks without the squashing function in the output layer are universal approximators. This result has been often used improperly because it has been applied to multilayer feedforward networks with the squashing function in the output layer. In this paper, we will prove that also this kind of neural networks are universal approximators, i.e. they are capable of approximating any Borel measurable function from one finite dimensional space into (0,1)(n) to any desired degree of accuracy, provided sufficiently many hidden units are available. (C) 2000 Elsevier Science Ltd. All rights reserved.
引用
收藏
页码：561 / 563
页数：3
相关论文
共 50 条
  • [1] Recurrent neural networks are universal approximators
    Schaefer, Anton Maximilian
    Zimmermann, Hans Georg
    [J]. ARTIFICIAL NEURAL NETWORKS - ICANN 2006, PT 1, 2006, 4131 : 632 - 640
  • [2] Recurrent neural networks are universal approximators
    Schaefer, Anton Maximilian
    Zimmermann, Hans-Georg
    [J]. INTERNATIONAL JOURNAL OF NEURAL SYSTEMS, 2007, 17 (04) : 253 - 263
  • [3] Layered neural networks as universal approximators
    Ciuca, I
    Ware, JA
    [J]. COMPUTATIONAL INTELLIGENCE: THEORY AND APPLICATIONS, 1997, 1226 : 411 - 415
  • [4] Radial basis function neural networks of Hankel translates as universal approximators
    Marrero, Isabel
    [J]. ANALYSIS AND APPLICATIONS, 2019, 17 (06) : 897 - 930
  • [5] Fuzzy continuous input-output controllers are universal approximators
    Wang, GJ
    [J]. FUZZY SETS AND SYSTEMS, 1998, 97 (01) : 95 - 99
  • [6] Recurrent Neural Networks Are Universal Approximators With Stochastic Inputs
    Chen, Xiuqiong
    Tao, Yangtianze
    Xu, Wenjie
    Yau, Stephen Shing-Toung
    [J]. IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS, 2023, 34 (10) : 7992 - 8006
  • [7] Spherical approximate identity neural networks are universal approximators
    Zainuddin, Zarita
    Fard, Saeed Panahian
    [J]. 2014 10TH INTERNATIONAL CONFERENCE ON NATURAL COMPUTATION (ICNC), 2014, : 72 - 76
  • [8] Toroidal Approximate Identity Neural Networks Are Universal Approximators
    Panahian Fard, Saeed
    Zainuddin, Zarita
    [J]. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 2014, 8834 : 135 - 142
  • [9] Toroidal Approximate Identity Neural Networks Are Universal Approximators
    Fard, Saeed Panahian
    Zainuddin, Zarita
    [J]. NEURAL INFORMATION PROCESSING (ICONIP 2014), PT I, 2014, 8834 : 135 - 142
  • [10] On function approximators implementable as layered neural networks
    Ciuca, I
    [J]. 24TH EUROMICRO CONFERENCE - PROCEEDING, VOLS 1 AND 2, 1998, : 663 - 669
12345