private NeuralNetworkModel(bool deserializing)
: base(deserializing)
{
if (deserializing)
{
multiLayerPerceptron = new alglib.multilayerperceptron();
}
}
public NeuralNetworkModel(alglib.multilayerperceptron multiLayerPerceptron, string targetVariable, IEnumerable<string> allowedInputVariables, double[] classValues = null)
: base(targetVariable) {
this.name = ItemName;
this.description = ItemDescription;
this.multiLayerPerceptron = multiLayerPerceptron;
this.allowedInputVariables = allowedInputVariables.ToArray();
if (classValues != null)
this.classValues = (double[])classValues.Clone();
}
public NeuralNetworkModel(alglib.multilayerperceptron multiLayerPerceptron, string targetVariable, IEnumerable <string> allowedInputVariables, double[] classValues = null)
: base(targetVariable)
{
this.name = ItemName;
this.description = ItemDescription;
this.multiLayerPerceptron = multiLayerPerceptron;
this.allowedInputVariables = allowedInputVariables.ToArray();
if (classValues != null)
{
this.classValues = (double[])classValues.Clone();
}
}
/// <summary>
/// Создает нейронную сеть с одним внутренним слоем
/// с границами и без взависимомти от флага границ
/// </summary>
/// <param name="bisLogObj">Имя объекта</param>
/// <param name="input">Количество входных нейронов</param>
/// <param name="edge">Флаг границ</param>
/// <param name="lowerEdge">Нижняя граница</param>
/// <param name="upperEdge">Верхняя граница</param>
private void Create(string bisLogObj, int input, bool edge, double lowerEdge, double upperEdge)
{
int layer = (int)(input * 1.2);
name = String.Format("{0}.{1}.{2}.{3}.{4}", bisLogObj, input, layer, lowerEdge, upperEdge);
net = new alglib.multilayerperceptron();
if (edge)
{
alglib.mlpcreater1(input, layer, 1, lowerEdge, upperEdge, out net);
}
else
{
alglib.mlpcreatec1(input, layer, 2, out net);
}
}
public static IRegressionSolution CreateNeuralNetworkRegressionSolution(IRegressionProblemData problemData, int nLayers, int nHiddenNodes1, int nHiddenNodes2, double decay, int restarts,
out double rmsError, out double avgRelError)
{
var dataset = problemData.Dataset;
string targetVariable = problemData.TargetVariable;
IEnumerable <string> allowedInputVariables = problemData.AllowedInputVariables;
IEnumerable <int> rows = problemData.TrainingIndices;
double[,] inputMatrix = dataset.ToArray(allowedInputVariables.Concat(new string[] { targetVariable }), rows);
if (inputMatrix.Cast <double>().Any(x => double.IsNaN(x) || double.IsInfinity(x)))
{
throw new NotSupportedException("Neural network regression does not support NaN or infinity values in the input dataset.");
}
alglib.multilayerperceptron multiLayerPerceptron = null;
if (nLayers == 0)
{
alglib.mlpcreate0(allowedInputVariables.Count(), 1, out multiLayerPerceptron);
}
else if (nLayers == 1)
{
alglib.mlpcreate1(allowedInputVariables.Count(), nHiddenNodes1, 1, out multiLayerPerceptron);
}
else if (nLayers == 2)
{
alglib.mlpcreate2(allowedInputVariables.Count(), nHiddenNodes1, nHiddenNodes2, 1, out multiLayerPerceptron);
}
else
{
throw new ArgumentException("Number of layers must be zero, one, or two.", "nLayers");
}
alglib.mlpreport rep;
int nRows = inputMatrix.GetLength(0);
int info;
// using mlptrainlm instead of mlptraines or mlptrainbfgs because only one parameter is necessary
alglib.mlptrainlm(multiLayerPerceptron, inputMatrix, nRows, decay, restarts, out info, out rep);
if (info != 2)
{
throw new ArgumentException("Error in calculation of neural network regression solution");
}
rmsError = alglib.mlprmserror(multiLayerPerceptron, inputMatrix, nRows);
avgRelError = alglib.mlpavgrelerror(multiLayerPerceptron, inputMatrix, nRows);
return(new NeuralNetworkRegressionSolution(new NeuralNetworkModel(multiLayerPerceptron, targetVariable, allowedInputVariables), (IRegressionProblemData)problemData.Clone()));
}
private NeuralNetworkModel(NeuralNetworkModel original, Cloner cloner)
: base(original, cloner) {
multiLayerPerceptron = new alglib.multilayerperceptron();
multiLayerPerceptron.innerobj.chunks = (double[,])original.multiLayerPerceptron.innerobj.chunks.Clone();
multiLayerPerceptron.innerobj.columnmeans = (double[])original.multiLayerPerceptron.innerobj.columnmeans.Clone();
multiLayerPerceptron.innerobj.columnsigmas = (double[])original.multiLayerPerceptron.innerobj.columnsigmas.Clone();
multiLayerPerceptron.innerobj.derror = (double[])original.multiLayerPerceptron.innerobj.derror.Clone();
multiLayerPerceptron.innerobj.dfdnet = (double[])original.multiLayerPerceptron.innerobj.dfdnet.Clone();
multiLayerPerceptron.innerobj.neurons = (double[])original.multiLayerPerceptron.innerobj.neurons.Clone();
multiLayerPerceptron.innerobj.nwbuf = (double[])original.multiLayerPerceptron.innerobj.nwbuf.Clone();
multiLayerPerceptron.innerobj.structinfo = (int[])original.multiLayerPerceptron.innerobj.structinfo.Clone();
multiLayerPerceptron.innerobj.weights = (double[])original.multiLayerPerceptron.innerobj.weights.Clone();
multiLayerPerceptron.innerobj.x = (double[])original.multiLayerPerceptron.innerobj.x.Clone();
multiLayerPerceptron.innerobj.y = (double[])original.multiLayerPerceptron.innerobj.y.Clone();
allowedInputVariables = (string[])original.allowedInputVariables.Clone();
if (original.classValues != null)
this.classValues = (double[])original.classValues.Clone();
}
private NeuralNetworkModel(NeuralNetworkModel original, Cloner cloner)
: base(original, cloner)
{
multiLayerPerceptron = new alglib.multilayerperceptron();
multiLayerPerceptron.innerobj.chunks = (double[, ])original.multiLayerPerceptron.innerobj.chunks.Clone();
multiLayerPerceptron.innerobj.columnmeans = (double[])original.multiLayerPerceptron.innerobj.columnmeans.Clone();
multiLayerPerceptron.innerobj.columnsigmas = (double[])original.multiLayerPerceptron.innerobj.columnsigmas.Clone();
multiLayerPerceptron.innerobj.derror = (double[])original.multiLayerPerceptron.innerobj.derror.Clone();
multiLayerPerceptron.innerobj.dfdnet = (double[])original.multiLayerPerceptron.innerobj.dfdnet.Clone();
multiLayerPerceptron.innerobj.neurons = (double[])original.multiLayerPerceptron.innerobj.neurons.Clone();
multiLayerPerceptron.innerobj.nwbuf = (double[])original.multiLayerPerceptron.innerobj.nwbuf.Clone();
multiLayerPerceptron.innerobj.structinfo = (int[])original.multiLayerPerceptron.innerobj.structinfo.Clone();
multiLayerPerceptron.innerobj.weights = (double[])original.multiLayerPerceptron.innerobj.weights.Clone();
multiLayerPerceptron.innerobj.x = (double[])original.multiLayerPerceptron.innerobj.x.Clone();
multiLayerPerceptron.innerobj.y = (double[])original.multiLayerPerceptron.innerobj.y.Clone();
allowedInputVariables = (string[])original.allowedInputVariables.Clone();
if (original.classValues != null)
{
this.classValues = (double[])original.classValues.Clone();
}
}
public static IClassificationSolution CreateNeuralNetworkClassificationSolution(IClassificationProblemData problemData, int nLayers, int nHiddenNodes1, int nHiddenNodes2, double decay, int restarts,
out double rmsError, out double avgRelError, out double relClassError)
{
var dataset = problemData.Dataset;
string targetVariable = problemData.TargetVariable;
IEnumerable <string> allowedInputVariables = problemData.AllowedInputVariables;
IEnumerable <int> rows = problemData.TrainingIndices;
double[,] inputMatrix = AlglibUtil.PrepareInputMatrix(dataset, allowedInputVariables.Concat(new string[] { targetVariable }), rows);
if (inputMatrix.Cast <double>().Any(x => double.IsNaN(x) || double.IsInfinity(x)))
{
throw new NotSupportedException("Neural network classification does not support NaN or infinity values in the input dataset.");
}
int nRows = inputMatrix.GetLength(0);
int nFeatures = inputMatrix.GetLength(1) - 1;
double[] classValues = dataset.GetDoubleValues(targetVariable).Distinct().OrderBy(x => x).ToArray();
int nClasses = classValues.Count();
// map original class values to values [0..nClasses-1]
Dictionary <double, double> classIndices = new Dictionary <double, double>();
for (int i = 0; i < nClasses; i++)
{
classIndices[classValues[i]] = i;
}
for (int row = 0; row < nRows; row++)
{
inputMatrix[row, nFeatures] = classIndices[inputMatrix[row, nFeatures]];
}
alglib.multilayerperceptron multiLayerPerceptron = null;
if (nLayers == 0)
{
alglib.mlpcreatec0(allowedInputVariables.Count(), nClasses, out multiLayerPerceptron);
}
else if (nLayers == 1)
{
alglib.mlpcreatec1(allowedInputVariables.Count(), nHiddenNodes1, nClasses, out multiLayerPerceptron);
}
else if (nLayers == 2)
{
alglib.mlpcreatec2(allowedInputVariables.Count(), nHiddenNodes1, nHiddenNodes2, nClasses, out multiLayerPerceptron);
}
else
{
throw new ArgumentException("Number of layers must be zero, one, or two.", "nLayers");
}
alglib.mlpreport rep;
int info;
// using mlptrainlm instead of mlptraines or mlptrainbfgs because only one parameter is necessary
alglib.mlptrainlm(multiLayerPerceptron, inputMatrix, nRows, decay, restarts, out info, out rep);
if (info != 2)
{
throw new ArgumentException("Error in calculation of neural network classification solution");
}
rmsError = alglib.mlprmserror(multiLayerPerceptron, inputMatrix, nRows);
avgRelError = alglib.mlpavgrelerror(multiLayerPerceptron, inputMatrix, nRows);
relClassError = alglib.mlpclserror(multiLayerPerceptron, inputMatrix, nRows) / (double)nRows;
var problemDataClone = (IClassificationProblemData)problemData.Clone();
return(new NeuralNetworkClassificationSolution(new NeuralNetworkModel(multiLayerPerceptron, targetVariable, allowedInputVariables, problemDataClone.ClassValues.ToArray()), problemDataClone));
}
private NeuralNetworkModel(bool deserializing)
: base(deserializing) {
if (deserializing)
multiLayerPerceptron = new alglib.multilayerperceptron();
}
private NeuralNetworkModel(StorableConstructorFlag _) : base(_)
{
multiLayerPerceptron = new alglib.multilayerperceptron();
}
