954.txt

This complete matlab for neural network

发信人: daniel (飞翔鸟), 信区: DataMining
标  题: Re: 请教一个概念：归纳学习？
发信站: 南京大学小百合站 (Sun Jun  2 22:35:19 2002), 站内信件


【 在 mining (key) 的大作中提到: 】
:  从认知的角度来看，学习的分类有：归纳学习、演绎学习、类比学习等
: 从机器学习的角度来看，可分为：机械式学习、示例学习、类比学习、归纳学习等等。
: 我想提的一个问题是：
: 从应用的角度来看，基于归纳思想的机器学习是否都应该考学习模型的推广能力
: （generalization）？
sure. this is the key of inductive learning

: 神经网络在这方面有大量的论述，而其他归纳学习方法为什么没有？
if you trace previous literatures, you may find that there were lots of
publications on the generalization ability of other kinds of inductive
learning methods. Discussions on the generalization of Version Space could 
be found in almost every ML book, which was investigated since 1970s.
FOr decision trees, most literatures on generalization were published in 
the mid of 1980s to the begining of 1990s, you can find some of them from early 
issues of ML Journal. FOr Bayes, most on Naive Bayes were published in 1970s,
most on Bayesian network were published in 1990s and up to now. For lazy
learning methods, most were in 1980s and the begining of 1990s, some can 
be found in David Aha's special issue of ML Journal. Even for neural networks,
most were published at the end of 1980s and the begining of 1990s.


: 极大极小规则在数学上有没有严格的论证？


: 【 在 tyqqre (tyqqre) 的大作中提到: 】
: : 有两种类型的学习方法：归纳法和演绎法。
: : 演绎方法：根据延械墓嬖颍ㄌ跫