artificial-neural-network-master (1)
所属分类:人工智能/神经网络/深度学习
开发工具:Java
文件大小:8KB
下载次数:1
上传日期:2019-06-20 15:51:13
上 传 者:
A7800741CS
说明: 基于java语言的人工神经网路实现,人工智能。
(Implementation of Artificial Neural Network Based on Java Language)
文件列表:
.classpath (226, 2017-01-23)
.project (384, 2017-01-23)
src (0, 2017-01-23)
src\ActivationFunction.java (2251, 2017-01-23)
src\Connection.java (2111, 2017-01-23)
src\Constants.java (312, 2017-01-23)
src\Layer.java (656, 2017-01-23)
src\NeuralNetwork.java (10728, 2017-01-23)
src\Neuron.java (4473, 2017-01-23)
ArtificialNeuralNetwork
=======================
An implementation of a (initially) fully-connected, feed-forward artificial neural network written in Java utilizing iRPROP+ training algorithm.
##Implementation Details
###iRPROP+
iRPROP+, or improved Resilient Propogation with Weight Backtracking, is a first-order optimization algorithm used for supervised learning of artificial neural networks. Based on Rprop developed by Martin Riedmiller and Heinrich Braun in 1992, iRPROP+ was created by Christian Igel and Michael Hsken in 2000 as an improved variant.
The Rprop family of error optimization takes into account the sign of the partial derivative over each pattern, updating all weights independently.
####References
Rprop+: http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.21.1417
iRPROP+: http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.17.1332
近期下载者:
相关文件:
收藏者: