文件列表:

.svn (0, 2012-10-09)
.svn\all-wcprops (5537, 2012-10-09)
.svn\dir-prop-base (140, 2012-10-09)
.svn\entries (9844, 2012-10-09)
.svn\text-base (0, 2012-10-09)
.svn\text-base\SVMdata.m.svn-base (798, 2012-10-09)
.svn\text-base\arrayToDims.m.svn-base (1641, 2012-10-09)
.svn\text-base\batch_RBM_update.m.svn-base (4333, 2012-10-09)
.svn\text-base\batch_RBMc_update.m.svn-base (4961, 2012-10-09)
.svn\text-base\c_gen_feat.m.svn-base (8611, 2012-10-09)
.svn\text-base\c_unsuperv_offline_eval.m.svn-base (8858, 2012-10-09)
.svn\text-base\cdbn_unsupervised_fit_testscript.m.svn-base (760, 2012-10-09)
.svn\text-base\convnFast.m.svn-base (8899, 2012-10-09)
.svn\text-base\convnfft.m.svn-base (6402, 2012-10-09)
.svn\text-base\convnfft_install.m.svn-base (438, 2012-10-09)
.svn\text-base\crbm2D.m.svn-base (13829, 2012-10-09)
.svn\text-base\cvis_features.m.svn-base (422, 2012-10-09)
.svn\text-base\cvis_features4D.m.svn-base (1127, 2012-10-09)
.svn\text-base\dbnFit.m.svn-base (1711, 2012-10-09)
.svn\text-base\dbnPredict.m.svn-base (573, 2012-10-09)
.svn\text-base\dbn_unsupervised_fit.m.svn-base (1266, 2012-10-09)
.svn\text-base\dispims.m.svn-base (389, 2012-10-09)
.svn\text-base\experiment_set1.m.svn-base (4714, 2012-10-09)
.svn\text-base\experiment_set2.m.svn-base (3199, 2012-10-09)
.svn\text-base\experiments.m.svn-base (242, 2012-10-09)
.svn\text-base\getObjects.m.svn-base (1527, 2012-10-09)
.svn\text-base\getbatches.m.svn-base (1329, 2012-10-09)
.svn\text-base\ind2sub2.m.svn-base (1227, 2012-10-09)
.svn\text-base\inplaceprod.c.svn-base (3070, 2012-10-09)
.svn\text-base\interweave.m.svn-base (409, 2012-10-09)
.svn\text-base\logistic.m.svn-base (65, 2012-10-09)
.svn\text-base\multinomial_exp.m.svn-base (2815, 2012-10-09)
.svn\text-base\multinomial_exp_sample.m.svn-base (1974, 2012-10-09)
.svn\text-base\myrepmat.m.svn-base (476, 2012-10-09)
.svn\text-base\mysub2ind2D.m.svn-base (82, 2012-10-09)
.svn\text-base\mysub2ind3D.m.svn-base (117, 2012-10-09)
.svn\text-base\normalize.m.svn-base (155, 2012-10-09)
.svn\text-base\notes.txt.svn-base (2654, 2012-10-09)
.svn\text-base\nunique.m.svn-base (977, 2012-10-09)
... ...

This is a Python library for several types of deep belief networks. The convolutional RBM code is based on the 2009 ICML paper by Lee, Grosse, Ranganath and Ng, "Convolutional Deep Belief Networks for Scalable Unsupervised Learning of Hierarchical Representations". All implementations incorporate an option to train hidden unit biases using a sparsity criterion, as described in Lee, Ekanadham and Ng, "Sparse Deep Belief Net Model for Visual Area V2" (NIPS 2008). All RBM implementations also provide an option to treat visible units as either binary or gaussian. Training networks with gaussian visible units is a tricky dance of parameter-twiddling, but binary units seem quite stable in their learning and convergence properties. To try things out, download the MNIST digits dataset from curl http://yann.lecun.com/exdb/mnist/train-images-idx3-ubyte.gz | gunzip -c > train-images.ubyte curl http://yann.lecun.com/exdb/mnist/train-labels-idx1-ubyte.gz | gunzip -c > train-labels.ubyte then install Python OpenGL 2.0 (make sure you updated your graphic card driver). There could be other libraries you need to install. and install glumpy using pip install http://glumpy.googlecode.com/hg#egg=glumpy Then run the test with python test_rbm.py --labels train-labels.ubyte --images train-images.ubyte --batch-size 25 --l2 0.0001 --alpha 0.02 --momentum 0.5 --tau 1000 --sparsity 0.01 The learning parameters are squirrelly, but if things go right you should see a number of images show up on your screen that represent the "basis functions" that the network has learned when trying to auto-encode the images you are feeding it. Please send me comments, corrections, or bug fixes ! Leif Johnson

近期下载者：

相关文件：

评论：[我要评论] [举报此文件]

收藏者：