RNNTrainOptions (Stanford JavaNLP API)

java.lang.Object
- edu.stanford.nlp.sentiment.RNNTrainOptions

All Implemented Interfaces:: java.io.Serializable

public class RNNTrainOptions
extends java.lang.Object
implements java.io.Serializable

See Also:: Serialized Form

Field Summary

Fields
Modifier and Type	Field and Description
`int`	`adagradResetFrequency` How many epochs between resets of the adagrad learning rates.
`int`	`batchSize`
`int`	`debugOutputEpochs`
`int`	`epochs` Number of times through all the trees
`double`	`initialAdagradWeight` The value to set the learning rate for each parameter when initializing adagrad.
`java.lang.String`	`initialMatrixLogPath` If set, the initial matrices are logged to this location as a single file using SentimentModel.toString()
`double`	`learningRate`
`int`	`maxTrainTimeSeconds`
`int`	`nThreads`
`double`	`regClassification` Regularization cost for the classification matrices
`double`	`regTransformMatrix` Regularization cost for the transform matrix
`double`	`regTransformTensor` Regularization cost for the transform tensor
`double`	`regWordVector` Regularization cost for the word vectors
`double`	`scalingForInit`
`boolean`	`shuffleMatrices` Shuffle matrices when training.

Constructor Summary

Constructors
Constructor and Description

RNNTrainOptions()

Constructors
Constructor and Description
`RNNTrainOptions()`

Method Summary

All Methods Instance Methods Concrete Methods
Modifier and Type	Method and Description
`double`	`getClassWeight(int i)` The classWeights can be passed in as a comma separated list of weights using the -classWeights flag.
`int`	`setOption(java.lang.String[] args, int argIndex)`
`java.lang.String`	`toString()`

Methods inherited from class java.lang.Object
clone, equals, finalize, getClass, hashCode, notify, notifyAll, wait, wait, wait

- Field Detail
  - batchSize
```
public int batchSize
```
  - epochs
```
public int epochs
```
    Number of times through all the trees
  - debugOutputEpochs
```
public int debugOutputEpochs
```
  - maxTrainTimeSeconds
```
public int maxTrainTimeSeconds
```
  - learningRate
```
public double learningRate
```
  - scalingForInit
```
public double scalingForInit
```
  - regTransformMatrix
```
public double regTransformMatrix
```
    Regularization cost for the transform matrix
  - regClassification
```
public double regClassification
```
    Regularization cost for the classification matrices
  - regWordVector
```
public double regWordVector
```
    Regularization cost for the word vectors
  - initialAdagradWeight
```
public double initialAdagradWeight
```
    The value to set the learning rate for each parameter when initializing adagrad.
  - adagradResetFrequency
```
public int adagradResetFrequency
```
    How many epochs between resets of the adagrad learning rates. Set to 0 to never reset.
  - regTransformTensor
```
public double regTransformTensor
```
    Regularization cost for the transform tensor
  - shuffleMatrices
```
public boolean shuffleMatrices
```
    Shuffle matrices when training. Usually should be true. Set to false to compare training across different implementations, such as with the original Matlab version
  - initialMatrixLogPath
```
public java.lang.String initialMatrixLogPath
```
    If set, the initial matrices are logged to this location as a single file using SentimentModel.toString()
  - nThreads
```
public int nThreads
```
- Constructor Detail
  - RNNTrainOptions
```
public RNNTrainOptions()
```
- Method Detail
  - getClassWeight
```
public double getClassWeight(int i)
```
    The classWeights can be passed in as a comma separated list of weights using the -classWeights flag. If the classWeights are not specified, the value is assumed to be 1.0. classWeights only apply at train time; we do not weight the classes at all during test time.
  - toString
```
public java.lang.String toString()
```
    Overrides:
    
    toString in class java.lang.Object
  - setOption
```
public int setOption(java.lang.String[] args,
                     int argIndex)
```

Class RNNTrainOptions

Field Summary

Constructor Summary

Method Summary

Methods inherited from class java.lang.Object

Field Detail

batchSize

epochs

debugOutputEpochs

maxTrainTimeSeconds

learningRate

scalingForInit

regTransformMatrix

regClassification

regWordVector

initialAdagradWeight

adagradResetFrequency

regTransformTensor

shuffleMatrices

initialMatrixLogPath

nThreads

Constructor Detail

RNNTrainOptions

Method Detail

getClassWeight

toString

setOption