文章分类 -  Deep Learning

摘要:参考网页:http://deeplearning.stanford.edu/wiki/index.php/Exercise:Sparse_Coding前言: 使用IMAGES数据集,在10张图片上采样20000个patch块,学习特征并对特征进行可视化表示。实验流程: 1、采样20000patc... 阅读全文
posted @ 2014-12-08 17:23 dupuleng 阅读(625) 评论(0) 推荐(0)
摘要:参考网页:http://deeplearning.stanford.edu/wiki/index.php/Deriving_gradients_using_the_backpropagation_ideaBP算法求解梯度,即将重构误差根据权重依次从输出层向前传输到隐含层,具体过程如下:接下来介绍如何... 阅读全文
posted @ 2014-12-08 16:40 dupuleng 阅读(479) 评论(0) 推荐(0)
摘要:参考网页:http://deeplearning.stanford.edu/wiki/index.php/Sparse_Codinghttp://deeplearning.stanford.edu/wiki/index.php/Sparse_Coding:_Autoencoder_Interpret... 阅读全文
posted @ 2014-12-08 15:33 dupuleng 阅读(575) 评论(0) 推荐(0)
摘要:参考网页:http://deeplearning.stanford.edu/wiki/index.php/Data_Preprocessinghttp://deeplearning.stanford.edu/wiki/index.php/Whitening前言: 数据预处理在深度学习中非常重要,通... 阅读全文
posted @ 2014-12-02 17:16 dupuleng 阅读(438) 评论(0) 推荐(0)
摘要:参考网页:http://deeplearning.stanford.edu/wiki/index.php/Feature_extraction_using_convolutionhttp://deeplearning.stanford.edu/wiki/index.php/Pooling实验介绍: ... 阅读全文
posted @ 2014-12-01 22:56 dupuleng 阅读(404) 评论(0) 推荐(0)
摘要:参考网页: http://deeplearning.stanford.edu/wiki/index.php/Linear_Decoders http://deeplearning.stanford.edu/wiki/index.php/Exercise:Learning_color_featur... 阅读全文
posted @ 2014-11-30 17:35 dupuleng 阅读(213) 评论(0) 推荐(0)
摘要:参考网页:http://deeplearning.stanford.edu/wiki/index.php/Exercise:_Implement_deep_networks_for_digit_classification实验介绍:完成手写体识别,采用MNIST手写数据库,数字从0-9,训练样本6万... 阅读全文
posted @ 2014-11-27 15:15 dupuleng 阅读(714) 评论(0) 推荐(1)
摘要:参考资料:http://deeplearning.stanford.edu/wiki/index.php/Self-Taught_Learninghttp://deeplearning.stanford.edu/wiki/index.php/Exercise:Self-Taught_Learning... 阅读全文
posted @ 2014-11-24 21:34 dupuleng 阅读(451) 评论(0) 推荐(0)
摘要:上一节介绍了softmax regression的理论知识,这一节介绍它的具体实现。具体参考http://deeplearning.stanford.edu/wiki/index.php/Exercise:Softmax_Regression实验介绍: 完成手写体识别,采用MNIST手写数... 阅读全文
posted @ 2014-11-24 12:34 dupuleng 阅读(380) 评论(0) 推荐(0)
摘要:softmax regression是对logistic regression 的扩展,由于logistic regression只能处理二分类问题,而实际应用中多分类问题随处可见,如手写体数字识别问题,此时的类别为10。有些人可能会想,这类问题我们也可以把它转化为二分类问题,这里就要谈到softm... 阅读全文
posted @ 2014-11-24 10:58 dupuleng 阅读(584) 评论(0) 推荐(0)