action2entropy: Step 1 of Subtask Analysis: obtaining entropy sequences of...

In xytangtang/ProcData: Process Data Analysis

Description

action2entropy fit a recurrent-neural-network-based action prediction model to a set of action sequences action_seqs.

Usage

action2entropy(action_seqs, rnn_dim = 20, n_epoch = 50,
  step_size = 0.001, batch_size = 1, optimizer_name = "rmsprop",
  index_valid = 0.2, verbose = FALSE)

Arguments

action_seqs	a list of action sequences
rnn_dim	latent dimension of RNN
n_epoch	the number of training epochs.
step_size	the learning rate of optimizer.
batch_size	the batch size used in training.
optimizer_name	a character string specifying the optimizer to be used for training. Availabel options are "sgd", "rmsprop", "adadelta", and "adam".
index_valid	proportion of sequences used as the validation set or a vector of indices specifying the validation set.
verbose	logical. If TRUE, training progress is printed.

Value

action2entropy returns a list containing

entropy_seqs	a list of entropy sequences. The length of each entropy sequence is one less than that of the corresponding action sequence.
loss_history	a n_epoch by 2 matrix. The two columns contain the loss at the end of each epoch on the training set and the validation set, respectively.
rnn_dim	the latent dimension of the recurrent neural network
model_fit	a vector of class "raw". It is the serialized version of the trained keras model.
actions	a vector of the actions in action_seqs.
max_len	maximum length of the action sequences.

References

Wang, Z., Tang, X., Liu, J., and Ying, Z. (2020) Subtask analysis of process data through a predictive model. https://arxiv.org/abs/2009.00717

See Also

entropy2segment and segment2subtask for steps 2 and 3 of the subtask analysis procedure; subtask_analysis for the complete procedure.

xytangtang/ProcData documentation built on May 30, 2021, 1:21 a.m.