Convolutional deep belief network

<p>In <a href="/facts/Computer_science/lN3pEyPz">computer science</a>, a convolutional deep belief network (CDBN) is a type of <a href="/facts/Deep_learning/JLuwD3ea">deep</a> <a href="/facts/Artificial_neural_network/6V1jMlkx">artificial neural network</a> composed of multiple layers of <a href="/facts/Convolutional_neural_network/kHEdnmGU">convolutional</a> <a href="/facts/Restricted_Boltzmann_machine/D5afVGCs">restricted Boltzmann machines</a> stacked together.  Alternatively, it is a hierarchical <a href="/facts/Generative_model/JUgExNIP">generative model</a> for deep learning, which is highly effective in <a href="/facts/Image_processing/zenalNCf">image processing</a> and <a href="/facts/Object_recognition/cdLpak1I">object recognition</a>, though it has been used in other domains too.  The salient features of the model include the fact that it scales well to high-dimensional images and is translation-invariant.
</p><p>CDBNs use the technique of probabilistic <a href="/facts/Max_pooling/kHEdnmGU">max-pooling</a> to reduce the dimensions in higher layers in the network. Training of the network involves a pre-training stage accomplished in a <a href="/facts/Greedy_algorithm/alfsekco">greedy</a> layer-wise manner, similar to other <a href="/facts/Deep_belief_network/aPkAjLIU">deep belief networks</a>. Depending on whether the network is to be used for discrimination or generative tasks, it is then "fine tuned" or trained with either <a href="/facts/Backpropogation/lCsIdKHc">back-propagation</a> or the up–down algorithm (contrastive–divergence), respectively.
</p>

Convolutional deep belief network open-in-new

Convolutional deep belief network