| 共 66 篇文章 |
|
阅172 转1 评0 公众公开 14-06-12 21:06 |
第二次群内论文讨论活动——Hyperparameter Optimization and Bayesian Optimization这次活动的主题是 hyperparameter optimization 和 bayesian optimization，简单说就是如何更有效率地调参，或者。。 阅1718 转8 评0 公众公开 14-02-19 10:35 |
VS2010下如何配置CUDA4.2.经过各种努力，终于搞定一个CUDA工程。第二：右键单击.cu文件，选择属性，然后General->Item Type->CUDA C/C ,确定。第三：仍然右键单击.cu文件，选择属性，CUDA C/C ->common,进行相应配置，不过貌似默认已经配好了。i ){//依次验证检测到的设备是否支持CUDA.//显示检测到的设备支持的CUDA 版本。if(i == co... 阅130 转1 评0 公众公开 13-08-02 19:19 |
below can probably be handled by insane matrix operations for j = 1 : net.layers{l}.outputmaps // for each output map // create temp output map z = zeros(size(net.layers{l - 1}.a{1}) - [net.layers{l}.kernelsize - 1 net.layers{l}.kernelsize - 1 0]);end ... 阅398 转10 评0 公众公开 13-08-01 21:16 |
【面向代码】学习 Deep Learning（二）Deep Belief Nets(DBNs)dbn =dbnsetup(dbn, train_x, opts);//unfold dbn to nn nn = dbnunfoldtonn(dbn, 10);\DBN\dbnsetup.m.\DBN\rbmtrain.m 代码如下，说明都在注释里 论文参考：【1】Learning Deep Architectures for AI 以及 【2】A Practical Guide to Training Restri... 阅1919 转18 评0 公众公开 13-08-01 21:15 |
sparsityError = [zeros(size(nn.a{i},1),1) nn.nonSparsityPenalty * (-nn.sparsityTarget ./ pi (1 - nn.sparsityTarget) ./ (1 - pi))];[cpp] view plaincopyfor i = 1 : (nn.n - 1) if(nn.weightPenaltyL2>0) dW = nn.dW{i} nn.weightPenaltyL2 * nn.if(nn.momentum>0) nn.vW{i} = nn.momentum*nn.vW{i} ... 阅3950 转29 评0 公众公开 13-08-01 21:15 |
A Unified Architecture for Natural Language Processing: Deep Neural Networks with Multitask Learning.(someone and tell us if you need to read the 3-way RBM paper before the mcRBM one [I didn''t find it necessary, in fact the CVPR paper seemed easier to understand.]) [26] Dahl, G., Ranzato, M., Mohamed, A. and ... 阅238 转1 评0 公众公开 13-07-30 20:46 |
function [] = optimizeAutoencoderLBFGS(layersizes, datasetpath, ...finalObjective)% train a deep autoencoder with variable hidden sizes% layersizes : the sizes of the hidden layers.lnew = lnew layersizes(i) * layersizes(i 1);W{i} = reshape(theta(lold:lnew), layersizes(i 1), layersizes(i));lnew = lnew layersizes(i 1);l... 阅610 转4 评0 公众公开 13-07-29 19:41 |
function [cost, grad] = orthonormalICACost(theta, visibleSize, numFeatures, patches, epsilon)%orthonormalICACost - compute the cost and gradients for orthonormal ICA% (i.e. compute the cost ||Wx||_1 and its gradient) weightMatrix = reshape(theta, numFeatures, visibleSize); 阅250 转2 评0 公众公开 13-07-29 19:40 |
此时如果有许多训练样本的话，同样可以给出训练样本的平均相关的误差函数，一般该函数也称作是损失函数（Loss function）。梯度下降法是用来求函数值最小处的参数值，而牛顿法是用来求函数值为0处的参数值，这两者的目的初看是感觉有所不同，但是再仔细观察下牛顿法是求函数值为0时的情况，如果此时的函数是某个函数A的导数，则牛顿法也算是求... 阅240 转6 评0 公众公开 13-07-29 19:13 |
