基于极端学习机算法的学习_独家原创.doc
约30页DOC格式手机打开展开
基于极端学习机算法的学习_独家原创,基于极端学习机算法的学习11900字自己原创的毕业论文，已经通过校内系统检测，重复率低，仅在本站独家出售，大家放心下载使用摘要 人工智能是一门边沿学科，属于自然科学、社会科学、技术科学三向的交叉学科。半个世纪以来，作为一门新兴学科，人工智能一直在不停地发展，其涉及的学科越来越多，研究的范畴也越来越大，应用的领域也越来越...
内容介绍
此文档由会员 小丑88 发布
基于极端学习机算法的学习
11900字
自己原创的毕业论文，已经通过校内系统检测，重复率低，仅在本站独家出售，大家放心下载使用
摘要 人工智能是一门边沿学科，属于自然科学、社会科学、技术科学三向的交叉学科。半个世纪以来，作为一门新兴学科，人工智能一直在不停地发展，其涉及的学科越来越多，研究的范畴也越来越大，应用的领域也越来越广。在其众多学科中，机器学习的前馈神经网络以及极端学习机算法便是本文要研究的内容。
通常，前馈神经网络的学习速度远远慢于需求。这已经成为了在过去的几十年里其应用的一个主要瓶颈。可能的两个关键原因是：1）较为缓慢的基于梯度的学习算法被广泛用来训练神经网络，以及2）网络中的所有参数都使用这种学习算法反复调整。因此，有学者提出了极端学习机（ELM）算法。ELM是一种简单易用、有效的单隐层前馈神经网络（SLFNs）学习算法，其本质在于隐层是不需要调整。理论上讲，这种算法易于以极快的学习速度提供更好的泛化性能。
尽管ELM性能优于前馈神经网络，但依旧有着一些缺陷。为此，研究人员又提出了改进的ELM算法以弥补ELM的一些不足，例如增量极端学习机（I-ELM）、加强I-ELM（EI-ELM）、自适应增长极端学习机（AG-ELM）、基于微粒群算法改进的极端学习机（PSO-ELM）等等。实验结果表明，PSO-ELM比ELM拥有更好的泛化性能，I-ELM、EI-ELM可以处理广泛的神经元，AG-ELM有优于ELM的逼近能力。本文对极端学习机及各种改进的极端学习机进行研究学习，并将其应用到sinc函数的回归和UCI数据集中。
关键词：人工智能；前馈神经网络；极端学习机。
Learning based on ELM algorithms
Abstract Artificial intelligence is an edge discipline, which is an interdisciplinary course combining natural science, social science and technology science. As a new discipline, artificial intelligence has been developing for half a century. Meanwhile, more and more disciplines have been involved, larger and larger areas has been researched and wider and wider field has been applied. Among its large number of disciplines, feedforward neural networks of machine learning and the algorithms of extreme learning machine(ELM) are to be studied in this paper.
As usual, the learning speed of feedforward neural networks is in general far lower than required and it has been a major bottleneck in their applications for past decades. Two key reasons behind may be: 1) the slow gradient-based learning algorithms are extensively used to train neural networks, and 2) all the parameters of the networks are tuned iteratively by using such learning algorithms. Therefore, some scholars have proposed extreme learning machine algorithm. ELM is an easy-to-use and effective algorithm of single hidden-layer feedforward neural networks (SLFNs) and its essence is that the hidden-layers don’t need to be tuned.
Although ELM performs better than feedforward neural networks, it still has some drawbacks. To make up the drawbacks, the researchers also proposed some improved algorithms of ELM such as: incremental ELM(I-ELM), enhanced incremental ELM(EI-ELM), Extreme Learning Machine with Adaptive Growth of Hidden Nodes(AG-ELM), and Extreme Learning Machine with Particle Swarm Optimization(PSO-ELM) etc. Experimental results show that PSO-ELM has better generalization performance than ELM; I-ELM, EI-ELM can handle a wide range of neurons; AG-ELM has a better approximation ability than ELM. In this paper, we will learn ELM and various improved ELM, and apply them to the regression of the function sinc and the data sets of UCI.
Key words Artificial intelligence; feedforward neural networks; ELM; drawbacks.
目 录
第一章 绪论 1
1.1 神经网络的发展史 1
1.2 前馈神经网络和BP算法 2
1.2.1 前馈神经网络概述 2
1.2.2 BP算法 3
1.3 极端学习机ELM的提出 5
1.4 论文的结构安排 5
第二章 极端学习机（ELM）相关概念及其算法 6
2.1 极端学习机的学习理论 6
2.2 极端学习机 7
2.2.1 摩尔彭德罗斯广义逆 7
2.2.2 极端学习机算法 8
2.2.3 极端学习机的优缺点 10
第三章 ELM的改进算法 11
3.1 I-ELM 11
3.2 AG-ELM 12
3.3 PSO-ELM 14
3.3.1 微粒群算法（PSO） 14
3.3.2 PSO-ELM 15
第四章 极端学习机及其改进算法的应用 18
4.1 PSO-ELM的应用和UCI数据集的分类 18
4.1.1. PSO-ELM应用于SINC函数逼近 18
4.1.2 PSO的UCI数据集分类 19
4.2 ELM及其改进算法在UCI数据集上的应用 20
结论 24
致谢 25
参考文献 26
11900字
自己原创的毕业论文，已经通过校内系统检测，重复率低，仅在本站独家出售，大家放心下载使用
摘要 人工智能是一门边沿学科，属于自然科学、社会科学、技术科学三向的交叉学科。半个世纪以来，作为一门新兴学科，人工智能一直在不停地发展，其涉及的学科越来越多，研究的范畴也越来越大，应用的领域也越来越广。在其众多学科中，机器学习的前馈神经网络以及极端学习机算法便是本文要研究的内容。
通常，前馈神经网络的学习速度远远慢于需求。这已经成为了在过去的几十年里其应用的一个主要瓶颈。可能的两个关键原因是：1）较为缓慢的基于梯度的学习算法被广泛用来训练神经网络，以及2）网络中的所有参数都使用这种学习算法反复调整。因此，有学者提出了极端学习机（ELM）算法。ELM是一种简单易用、有效的单隐层前馈神经网络（SLFNs）学习算法，其本质在于隐层是不需要调整。理论上讲，这种算法易于以极快的学习速度提供更好的泛化性能。
尽管ELM性能优于前馈神经网络，但依旧有着一些缺陷。为此，研究人员又提出了改进的ELM算法以弥补ELM的一些不足，例如增量极端学习机（I-ELM）、加强I-ELM（EI-ELM）、自适应增长极端学习机（AG-ELM）、基于微粒群算法改进的极端学习机（PSO-ELM）等等。实验结果表明，PSO-ELM比ELM拥有更好的泛化性能，I-ELM、EI-ELM可以处理广泛的神经元，AG-ELM有优于ELM的逼近能力。本文对极端学习机及各种改进的极端学习机进行研究学习，并将其应用到sinc函数的回归和UCI数据集中。
关键词：人工智能；前馈神经网络；极端学习机。
Learning based on ELM algorithms
Abstract Artificial intelligence is an edge discipline, which is an interdisciplinary course combining natural science, social science and technology science. As a new discipline, artificial intelligence has been developing for half a century. Meanwhile, more and more disciplines have been involved, larger and larger areas has been researched and wider and wider field has been applied. Among its large number of disciplines, feedforward neural networks of machine learning and the algorithms of extreme learning machine(ELM) are to be studied in this paper.
As usual, the learning speed of feedforward neural networks is in general far lower than required and it has been a major bottleneck in their applications for past decades. Two key reasons behind may be: 1) the slow gradient-based learning algorithms are extensively used to train neural networks, and 2) all the parameters of the networks are tuned iteratively by using such learning algorithms. Therefore, some scholars have proposed extreme learning machine algorithm. ELM is an easy-to-use and effective algorithm of single hidden-layer feedforward neural networks (SLFNs) and its essence is that the hidden-layers don’t need to be tuned.
Although ELM performs better than feedforward neural networks, it still has some drawbacks. To make up the drawbacks, the researchers also proposed some improved algorithms of ELM such as: incremental ELM(I-ELM), enhanced incremental ELM(EI-ELM), Extreme Learning Machine with Adaptive Growth of Hidden Nodes(AG-ELM), and Extreme Learning Machine with Particle Swarm Optimization(PSO-ELM) etc. Experimental results show that PSO-ELM has better generalization performance than ELM; I-ELM, EI-ELM can handle a wide range of neurons; AG-ELM has a better approximation ability than ELM. In this paper, we will learn ELM and various improved ELM, and apply them to the regression of the function sinc and the data sets of UCI.
Key words Artificial intelligence; feedforward neural networks; ELM; drawbacks.
目 录
第一章 绪论 1
1.1 神经网络的发展史 1
1.2 前馈神经网络和BP算法 2
1.2.1 前馈神经网络概述 2
1.2.2 BP算法 3
1.3 极端学习机ELM的提出 5
1.4 论文的结构安排 5
第二章 极端学习机（ELM）相关概念及其算法 6
2.1 极端学习机的学习理论 6
2.2 极端学习机 7
2.2.1 摩尔彭德罗斯广义逆 7
2.2.2 极端学习机算法 8
2.2.3 极端学习机的优缺点 10
第三章 ELM的改进算法 11
3.1 I-ELM 11
3.2 AG-ELM 12
3.3 PSO-ELM 14
3.3.1 微粒群算法（PSO） 14
3.3.2 PSO-ELM 15
第四章 极端学习机及其改进算法的应用 18
4.1 PSO-ELM的应用和UCI数据集的分类 18
4.1.1. PSO-ELM应用于SINC函数逼近 18
4.1.2 PSO的UCI数据集分类 19
4.2 ELM及其改进算法在UCI数据集上的应用 20
结论 24
致谢 25
参考文献 26
