com.structed.models

Class StructEDModel

java.lang.Object

com.structed.models.StructEDModel

All Implemented Interfaces:

java.io.Serializable

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

Created by yossiadi on 6/28/15. The basic model class.

See Also:

Serialized Form

Constructor Summary

Constructors

Constructor and Description
StructEDModel(Vector init_weights, IUpdateRule updateRule, ITaskLoss taskLoss, IInference inference, IKernel kernel, IFeatureFunctions phi, java.util.ArrayList<java.lang.Double> args)
StructEDModel(Vector init_weights, IUpdateRule updateRule, ITaskLoss taskLoss, IInference inference, IKernel kernel, IFeatureFunctions phi, java.util.ArrayList<java.lang.Double> args, boolean isShuffle)

Method Summary

Methods

Modifier and Type	Method and Description
double	getCumulative_loss()
Vector	getWeights()
void	loadModel(java.lang.String path) load the model, notice: we load only the weight vector, the rest of the configuration should be predefined
void	plotValidationError(boolean save) Plotting the error graph for the validation set
java.util.ArrayList<PredictedLabels>	predict(InstancesContainer instances, java.util.ArrayList<java.lang.Double> task_loss_params, int numPredictions2Return, boolean verbose) predict based on the model and return the scores of the best matches
void	saveModel(java.lang.String path) save the model, notice: we save only the weight vector, the rest of the configuration should be predefined
void	setFeatureFunctions(IFeatureFunctions phi)
void	setInference(IInference inference)
void	setKernel(IKernel kernel)
void	setReShuffle(boolean isShuffle)
void	setTaskLoss(ITaskLoss taskLoss)
void	setUpdateRule(IUpdateRule updateRule)
void	train(InstancesContainer trainInstances, java.util.ArrayList<java.lang.Double> task_loss_params, InstancesContainer developInstances, int epoch, int isAvg, boolean verbose) Train the model on the train instances

Methods inherited from class java.lang.Object

equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

StructEDModel

public StructEDModel(Vector init_weights,
             IUpdateRule updateRule,
             ITaskLoss taskLoss,
             IInference inference,
             IKernel kernel,
             IFeatureFunctions phi,
             java.util.ArrayList<java.lang.Double> args)

StructEDModel

public StructEDModel(Vector init_weights,
             IUpdateRule updateRule,
             ITaskLoss taskLoss,
             IInference inference,
             IKernel kernel,
             IFeatureFunctions phi,
             java.util.ArrayList<java.lang.Double> args,
             boolean isShuffle)

Method Detail

train

public void train(InstancesContainer trainInstances,
         java.util.ArrayList<java.lang.Double> task_loss_params,
         InstancesContainer developInstances,
         int epoch,
         int isAvg,
         boolean verbose)
           throws java.lang.Exception

Train the model on the train instances

Parameters:

epoch - - the number of epochs to run over the data

trainInstances - - the training instances(train set)

task_loss_params - - the task loss parameters, if there are no parameters set this to null

developInstances - - the develop instances(dev set), set this to null if there is no validation set

isAvg - - an indicator whether or not to average the weights

Throws:

java.lang.Exception

predict

public java.util.ArrayList<PredictedLabels> predict(InstancesContainer instances,
                                           java.util.ArrayList<java.lang.Double> task_loss_params,
                                           int numPredictions2Return,
                                           boolean verbose)

predict based on the model and return the scores of the best matches

Parameters:

instances - - the test set

task_loss_params - - the parameters for the task loss function, set to null if there are no task loss parameters

numPredictions2Return - - the number of examples to return in the scores

Returns:

ArrayList - return the predicted labels as array

plotValidationError

public void plotValidationError(boolean save)

Plotting the error graph for the validation set

Parameters:

save - flag which indicates whether to write the image or not

saveModel

public void saveModel(java.lang.String path)
               throws java.io.IOException

save the model, notice: we save only the weight vector, the rest of the configuration should be predefined

Parameters:

path - - the path to save the model

Throws:

java.io.IOException

loadModel

public void loadModel(java.lang.String path)
               throws java.io.IOException,
                      java.lang.ClassNotFoundException

load the model, notice: we load only the weight vector, the rest of the configuration should be predefined

Parameters:

path - - the model's path

Throws:

java.io.IOException

java.lang.ClassNotFoundException

getWeights

public Vector getWeights()

getCumulative_loss

public double getCumulative_loss()

setUpdateRule

public void setUpdateRule(IUpdateRule updateRule)

setTaskLoss

public void setTaskLoss(ITaskLoss taskLoss)

setKernel

public void setKernel(IKernel kernel)

setInference

public void setInference(IInference inference)

setFeatureFunctions

public void setFeatureFunctions(IFeatureFunctions phi)

setReShuffle

public void setReShuffle(boolean isShuffle)