Returns a List which may be used as model.par
of e.g. the function Setup.QLearning()
with the following parameters:
name - Identifier of Model. Per Default \"RNN\".
setup - Function which should be used to setup the RNN. Per Default Setup.RNN
predict - Function which should be used to predict the RNN. Per Default Predict.RNN
train - Function which should be used to train/calibrate the RNN. Per Default Train.RNN
hidden.nodes - A Vector consisting of the number of Neurons in each hidden layer - e.g. c(25,10) to have two hidden layers with the first layer having 25 Neurons.
layer.type - A vector consisting of the names of the type of the hidden layer. Supported are "lstm", "gru", "dense". If lstm or gru are used in a deep layer the sequence is returned.
activation.hidden - A Vector defining the activation functions of the hidden layers, e.g. c(\"relu\",\"relu\"). Has to have the same number of items as hidden.nodes
. Supported are e.g. relu, tanh, sigmoid and linear
activation.output. Activiation function of the output layer. Supported are e.g. relu, tanh, sigmoid and linear.
loss. Specifies the loss function, e.g. \'mse\'
optimizer. Specifies the used optimizer. By Default Adam Optimization is used with a Learning rate of 0.001.
mask.value. Which value should be used for masking?
epochs. How many epochs should the RNN be trained?
batch.size. Batch Size of RNN.
verbose. Should the RNN give an output? 0 for no output, 1 for output for each epoch, 2 for aggregate output every other epoch.
1 | Get.Def.Par.RNN(setting = "ThesisOpt")
|
setting |
Specefies the to be used setting. Currently the following settings are available:
|
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.