An implementation of deep neural network with rectifier linear units trained with stochastic gradient descent method and batch normalization. A combination of these methods have achieved state-of-the-art performance in ImageNet classification by overcoming the gradient saturation problem experienced by many deep architecture neural network models in the past. In addition, batch normalization and dropout are implemented as a means of regularization. The deeplearning package is inspired by the darch package and uses its class DArch.
|Author||Zhi Ruan [aut, cre], Martin Drees [cph]|
|Date of publication||2016-04-11 18:09:06|
|Maintainer||Zhi Ruan <firstname.lastname@example.org>|
|License||GPL (>= 2)|
applyDropoutMask: Applies the given dropout mask to the given data row-wise.
AR: Calculates the Accuracy Ratio of a classifier
AR.DArch: Calculates the Accruacy Ratio of a given set of probability
AR.default: Calculates the Accruacy Ratio of a given set of probability
AR.numeric: Calculates the Accruacy Ratio of a given set of probability
backpropagate_delta_bn: Calculates the delta functions using backpropagation
batch_normalization: Batch Normalization Function that normalizes the input before...
batch_normalization_differential: Function that calcualtes the differentials in the batch...
calcualte_population_mu_sigma: Calculates the mu and sigmas of a darch instance
classification_error: Calculates the classification error
convert_categorical: Data proprosess function that covnerts a categorical input to...
crossEntropyErr: Calculates the cross entropy error
finetune_SGD_bn: Updates a deep neural network's parameters using stochastic...
generateDropoutMask: Generates the dropout mask for the deep neural network
generateDropoutMasksForDarch: Generates dropout masks for dnn
matMult: Calculates the outer product of two matricies
meanSquareErr: Calculates the mean squared error
new_dnn: Creats a new instance of darch class
print_weight: Prints out the weight of a deep neural network
rectified_linear_unit_function: Rectified Linear Unit Function
reset_population_mu_sigma: Resets the mu and sigmas of a darch instance to 0 and 1
rsq: Calculate the RSQ of a regression model Utilitiy function...
rsq.DArch: Utilitiy function that calcualtes RSQ of a DArch instance
rsq.lm: Utilitiy function that calcualtes RSQ of a linear model
run_dnn: Execution function that runs in the batch normalization mode
train_dnn: Train a deep neural network
verticalize: Creates a matrix by repeating a row vector N times