CharRNN

Full name Microsoft.VisualBasic.MachineLearning.RNN.CharRNN Assembly Microsoft.VisualBasic.DeepLearning Members 5

Command line helpers for training, sampling and snapshotting a character level recurrent network.

00 Remarks

https://github.com/garstka/char-rnn-java

01 Syntax

Microsoft.VisualBasic.MachineLearning.RNN.CharRNN

02 Methods

Name	Overloads	Summary
initialize	1	Creates a network configured from the given options: a single layer network when Options.useSingleLayerNet is set, otherwise a multi layer network with the requested layer count.
train	1	Trains the network on the data set referenced by the options, sampling and snapshotting at the configured intervals.
saveASnapshot	1	Saves a network snapshot under the given name.
loadASnapshot	1	Loads a network snapshot with the given name.
sample	1	Samples the network for n characters and prints the result to the console.

03 Members

method initialize #

initialize(Options)

Creates a network configured from the given options: a single layer network when Options.useSingleLayerNet is set, otherwise a multi layer network with the requested layer count.

Parameters

Name	Type	Description
`options`	Options	The application options that describe the network.

Returns

The new, not yet initialized network.

method train #

train(Options, CharLevelRNN, String)

Trains the network on the data set referenced by the options, sampling and snapshotting at the configured intervals.

Parameters

Name	Type	Description
`options`	Options	The training options.
`net`	CharLevelRNN	The network to train.
`snapshotName`	`String`	Base name used for the snapshot files.

method saveASnapshot #

saveASnapshot(String, CharLevelRNN)

Saves a network snapshot under the given name.

Parameters

Name	Type	Description
`name`	`String`	Base name of the snapshot file.
`net`	CharLevelRNN	The network to snapshot.

method loadASnapshot #

loadASnapshot(String)

Loads a network snapshot with the given name.

Parameters

Name	Type	Description
`name`	`String`	Base name of the snapshot file.

Returns

The restored network.

method sample #

sample(Int32, String, Double, CharLevelRNN)

Samples the network for n characters and prints the result to the console.

Parameters

Name	Type	Description
`n`	`Int32`	Number of characters to sample; it must be at least 1.
`seed`	`String`	The seed text; it must not be `Nothing`.
`temperature`	`Double`	Sampling temperature in `(0.0, 1.0]`.
`net`	CharLevelRNN	The network to sample from; it must be initialized.