本文详细介绍了受限玻尔兹曼机（RBM）的训练过程，包括对比差异（CD）方法的回顾、统计收集、mini-batch大小、学习率、权重和偏置的初始化、动量、权衰减、刺激稀疏隐藏激活、隐藏层单元数量、不同类型的单元，以及对比差异的多样性，提供了一套实用的指导原则。
本栏目（Training Restricted Boltzmann Machines）包括内容的介绍、局限玻尔斯曼机与对比差异（CD）的回顾、使用CD时怎样纠正统计、迷你块大小（mini-batch)、检查学习的过程，检查过拟合，学习率、权值和偏置的初始值、能量、权值衰减、刺激稀疏隐藏激活、隐藏层单元数
、单元不同的类型等章节。所有内容均来自Geoffrey Hinton大师的这篇文章 A Practical Guide to Training Restricted Boltzmann Machines。（http://learning.cs.toronto.edu
）
讲义的目录：
2 An overview of Restricted Boltzmann Machines and Contrastive Divergence
3 How to collect statistics when using Contrastive Divergence
4 The size of a mini-batch
5 Monitoring the progress of learning
6 Monitoring the over tting
7 The learning rate
8 The initial values of the weights and biases
9 Momentum
10 Weight-decay
11 Encouraging sparse hidden activities
12 The number of hidden units
13 Diff erent types of unit
4 The size of a mini-batch
5 Monitoring the progress of learning
6 Monitoring the over tting
7 The learning rate
8 The initial values of the weights and biases
9 Momentum
10 Weight-decay
11 Encouraging sparse hidden activities
12 The number of hidden units
13 Diff erent types of unit
14 Varieties of contrastive divergence
15 Displaying what is happening during learning
16 Using RBM's for discrimination
17 Dealing with missing values
17 Dealing with missing values
扫一扫
1340
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
