Prior to connecting the output of the (second) convolution layer with a fully connected network, in step 1, we reshape the four-dimensional convolution layer into a two-dimensional tensor. The first dimension (?) represents any number of input images (as rows) and the second dimension represents the flattened vector of features generated for each image of length 4,096; that is, 8 x 8 x 64 (as columns). Steps 2 and 3 validate the dimensions of the reshaped layers and input features.
- Preface
- Getting Started
- Deep Learning with R
- Starting with logistic regression
- Introducing the dataset
- Performing logistic regression using H2O
- Performing logistic regression using TensorFlow
- Visualizing TensorFlow graphs
- Starting with multilayer perceptrons
- Setting up a neural network using H2O
- Tuning hyper-parameters using grid searches in H2O
- Setting up a neural network using MXNet
- Setting up a neural network using TensorFlow
- Convolution Neural Network
- Introduction
- Downloading and configuring an image dataset
- Learning the architecture of a CNN classifier
- Using functions to initialize weights and biases
- Using functions to create a new convolution layer
- Using functions to create a new convolution layer
- Using functions to flatten the densely connected layer
- Defining placeholder variables
- Creating the first convolution layer
- Creating the second convolution layer
- Flattening the second convolution layer
- Creating the first fully connected layer
- Applying dropout to the first fully connected layer
- Creating the second fully connected layer with dropout
- Applying softmax activation to obtain a predicted class
- Defining the cost function used for optimization
- Performing gradient descent cost optimization
- Executing the graph in a TensorFlow session
- Evaluating the performance on test data
- Data Representation Using Autoencoders
- Introduction
- Setting up autoencoders
- Data normalization
- Setting up a regularized autoencoder
- Fine-tuning the parameters of the autoencoder
- Setting up stacked autoencoders
- Setting up denoising autoencoders
- Building and comparing stochastic encoders and decoders
- Learning manifolds from autoencoders
- Evaluating the sparse decomposition
- Generative Models in Deep Learning
- Comparing principal component analysis with the Restricted Boltzmann machine
- Setting up a Restricted Boltzmann machine for Bernoulli distribution input
- Training a Restricted Boltzmann machine
- Backward or reconstruction phase of RBM
- Understanding the contrastive divergence of the reconstruction
- Initializing and starting a new TensorFlow session
- Evaluating the output from an RBM
- Setting up a Restricted Boltzmann machine for Collaborative Filtering
- Performing a full run of training an RBM
- Setting up a Deep Belief Network
- Implementing a feed-forward backpropagation Neural Network
- Setting up a Deep Restricted Boltzmann Machine
- Recurrent Neural Networks
- Reinforcement Learning
- Application of Deep Learning in Text Mining
- Application of Deep Learning to Signal processing
- Transfer Learning
How it works...
-
No Comment
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.