DLflow.train: Train Part of a Flow
In neuroimaginador/dl4ni: Deep Learning Flows for Neuroimaging

This function is used to train a DLmodel that specifies an output of the flow.

`output`	(charcater) Name of the output (model) to train.
`input_filenames`	(list) List of input files to train with, that is, the X in the training phase.
`output_filenames`	(list) List of output files, representing the Y in the training phase.
`train_split`	(numeric) Ratio of input files used for training, the remaining are used for validation, Default: 0.75
`epochs`	(numeric) Number of epochs to train for, Default: 10
`target_windows_per_file`	(numeric) Number of windows to extract, as a minimum, per file, Default: 1024
`mode`	(character) Inference mode for the required inputs of this part of the flow, in case they had to be computed, Default: c("debug", "faster", "medium", "slower")

In case the part of the flow is currently defined by a DLscheme, a DLmodel is instatiated taking into account given inputs and output filenames. This is the model to be trained.

The flow with the trained model.

neuroimaginador/dl4ni documentation built on May 3, 2019, 5:47 p.m.