Deep Learning

Deep learning is often used to attempt to automatically learn representations of data with multiple layers of information-processing modules in hierarchical architectures. This chapter presents several deep neural techniques and their applications in machine fault diagnosis. These include autoencoder-based deep neural networks, convolutional neural networks (CNNs), deep belief networks (DBNs), and recurrent neural networks (RNNs). Unlike a standard neural network (NN), the architecture of a CNN is usually composed of a convolutional layer and a subsampling layer also called a pooling layer. DBNs are generative NNs that stack multiple restricted Boltzmann machines that can be trained in a greedy layer-wise unsupervised way; then they can be further fine-tuned with respect to labels of training data by adding a softmax layer in the top layer. As is the case with DBNs, RNNs can be trained via backpropagation through time for supervised tasks with sequential input data and target outputs.