encog(二)神经网络的训练

Encog中有很多的训练方法。

EncogUtility是一个功能辅助类，提供了很多方便的函数

Modifier and Type	Method and Description
static double	calculateClassificationError(MLClassification method, MLDataSet data) Calculate the classification error.
static double	calculateRegressionError(MLRegression method, MLDataSet data)
static void	convertCSV2Binary(File csvFile, CSVFormat format, File binFile, int[] input, int[] ideal, boolean headers)
static void	convertCSV2Binary(File csvFile, File binFile, int inputCount, int outputCount, boolean headers) Convert a CSV file to a binary training file.
static void	convertCSV2Binary(String csvFile, String binFile, int inputCount, int outputCount, boolean headers) Convert a CSV file to a binary training file.
static void	evaluate(MLRegression network, MLDataSet training) Evaluate the network and display (to the console) the output for every value in the training set.
static void	explainErrorMSE(MLRegression method, MatrixMLDataSet training)
static void	explainErrorRMS(MLRegression method, MatrixMLDataSet training)
static String	formatNeuralData(MLData data) Format neural data as a list of numbers.
static MLDataSet	loadCSV2Memory(String filename, int input, int ideal, boolean headers, CSVFormat format, boolean significance) Load CSV to memory.
static MLDataSet	loadEGB2Memory(File filename)
static void	saveCSV(File targetFile, CSVFormat format, MLDataSet set)
static void	saveEGB(File f, MLDataSet data) Save a training set to an EGB file.
static BasicNetwork	simpleFeedForward(int input, int hidden1, int hidden2, int output, boolean tanh) Create a simple feedforward neural network.
static void	trainConsole(BasicNetwork network, MLDataSet trainingSet, int minutes) Train the neural network, using SCG training, and output status to the console.
static void	trainConsole(MLTrain train, BasicNetwork network, MLDataSet trainingSet, int minutes) Train the network, using the specified training algorithm, and send the output to the console.
static void	trainToError(MLMethod method, MLDataSet dataSet, double error) Train the method, to a specific error, send the output to the console.
static void	trainToError(MLTrain train, double error) Train to a specific error, using the specified training method, send the output to the console.

BasicTraining类是所有训练方法类的父类

构造函数

BasicTraining()

Used for serialization.

BasicTraining(TrainingImplementationType implementationType) 

 

返回值	成员函数
void	addStrategy(Strategy strategy) Training strategies can be added to improve the training results.
void	finishTraining() Should be called after training has completed and the iteration method will not be called any further.
double	getError()
TrainingImplementationType	getImplementationType()
int	getIteration()
List<Strategy>	getStrategies()
MLDataSet	getTraining()
boolean	isTrainingDone()
void	iteration(int count) Perform the specified number of training iterations.
void	postIteration() Call the strategies after an iteration.
void	preIteration() Call the strategies before an iteration.
void	setError(double error)
void	setIteration(int iteration) Set the current training iteration.
void	setTraining(MLDataSet training) Set the training object that this strategy is working with.

Backpropagation类是propagation类的子类

构造函数

Backpropagation(ContainsFlat network, MLDataSet training)

Create a class to train using backpropagation.

Backpropagation(ContainsFlat network, MLDataSet training, double theLearnRate, double theMomentum) 

第一个参数：将被训练的网络

第二个参数: 训练集

第三个参数：学习率

第四个参数: 梯度下降法中一种常用的加速技术。momentum是加速系数，momentum=0表示无加速，值越大表示加速越快。

 

返回值	成员函数
boolean	canContinue()
double[]	getLastDelta()
double	getLearningRate()
double	getMomentum()
void	initOthers() Perform training method specific init.
boolean	isValidResume(TrainingContinuation state) Determine if the specified continuation object is valid to resume with.
TrainingContinuation	pause() Pause the training.
void	resume(TrainingContinuation state) Resume training.
void	setLearningRate(double rate) Set the learning rate, this is value is essentially a percent.
void	setMomentum(double m) Set the momentum for training.
double	updateWeight(double[] gradients, double[] lastGradient, int index) Update a weight.