public static DBeta FromMatlabStruct(MatlabStruct s)
{
string className = s.GetString("className");
if (!className.Equals(MATLAB_CLASS))
{
throw new ArgumentException("The input does not represent a " + typeof(DBeta));
}
double a = s.GetDouble("alpha");
double b = s.GetDouble("beta");
return(new DBeta(a, b));
}
public new static KGGaussian <T> FromMatlabStruct(MatlabStruct s)
{
// s = struct();
// s.className=class(this);
// s.kegauss = this.kegauss.toStruct();
// s.embed_width2 = this.embed_width2;
// s.width2 = this.width2;
//
string className = s.GetString("className");
if (!className.Equals(MATLAB_CLASS))
{
throw new ArgumentException("The input does not represent a " + typeof(KGGaussian <T>));
}
// KEGaussian<T> keGauss = KEGaussian<T>.FromMatlabStruct(s.GetStruct("kegauss"));
double[] embedSquaredWidths = s.Get1DDoubleArray("embed_width2s");
if (!MatrixUtils.IsAllPositive(embedSquaredWidths))
{
throw new ArgumentException("all embedding width^2's must be positive.");
}
double squaredWidth = s.GetDouble("width2");
if (squaredWidth <= 0)
{
throw new ArgumentException("width2 must be > 0");
}
return(new KGGaussian <T>(embedSquaredWidths, squaredWidth));
}
public new static RFGJointKGG FromMatlabStruct(MatlabStruct s)
{
// s.className=class(this);
// s.embed_width2s_cell = this.embed_width2s_cell;
// s.outer_width2 = this.outer_width2;
// s.numFeatures=this.numFeatures;
// s.innerNumFeatures = this.innerNumFeatures;
// s.eprodMap=this.eprodMap.toStruct();
// s.Wout = this.Wout;
// s.Bout = this.Bout;
string className = s.GetString("className");
if (!className.Equals(MATLAB_CLASS))
{
throw new ArgumentException("The input does not represent a " + MATLAB_CLASS);
}
double outer_width2 = s.GetDouble("outer_width2");
int numFeatures = s.GetInt("numFeatures");
int innerNumFeatures = s.GetInt("innerNumFeatures");
MatlabStruct mapStruct = s.GetStruct("eprodMap");
RFGEProdMap eprodMap = RFGEProdMap.FromMatlabStruct(mapStruct);
Matrix Wout = s.GetMatrix("Wout");
if (innerNumFeatures != Wout.Rows)
{
throw new ArgumentException("inner #features must be = #rows of Wout");
}
if (numFeatures != Wout.Cols)
{
throw new ArgumentException("numFeatures must be = #cols of Wout");
}
Vector Bout = s.Get1DVector("Bout");
if (Bout.Count != numFeatures)
{
throw new ArgumentException("Bout must have length = numFeatures");
}
RFGJointKGG jointMap = new RFGJointKGG();
jointMap.outer_width2 = outer_width2;
jointMap.numFeatures = numFeatures;
jointMap.innerNumFeatures = innerNumFeatures;
jointMap.eprodMap = eprodMap;
jointMap.Wout = Wout;
jointMap.Bout = Bout;
// construct object
return(jointMap);
}
public static new BayesLinRegFM FromMatlabStruct(MatlabStruct s)
{
// s.className=class(this);
// s.featureMap=this.featureMap.toStruct();
// %s.regParam=this.regParam;
// s.mapMatrix=this.mapMatrix;
// s.posteriorCov = this.posteriorCov;
// s.noise_var = this.noise_var;
string className = s.GetString("className");
if (!className.Equals(MATLAB_CLASS))
{
throw new ArgumentException("The input does not represent a " + MATLAB_CLASS);
}
MatlabStruct fmStruct = s.GetStruct("featureMap");
RandomFeatureMap featureMap = RandomFeatureMap.FromMatlabStruct(fmStruct);
// This is the same as a posterior mean
Vector mapMatrix = s.Get1DVector("mapMatrix");
if (mapMatrix.Count != featureMap.GetOutputDimension())
{
throw new ArgumentException("mapMatrix and featureMap's dimenions are incompatible.");
}
Matrix postCov = s.GetMatrix("posteriorCov");
if (postCov.Cols != featureMap.GetOutputDimension())
{
throw new ArgumentException("posterior covariance and featureMap's dimenions are incompatible.");
}
double noise_var = s.GetDouble("noise_var");
Vector crossCorr = s.Get1DVector("crossCorrelation");
var bayes = new BayesLinRegFM();
bayes.featureMap = featureMap;
bayes.posteriorMean = mapMatrix;
bayes.posteriorCov = postCov;
bayes.noiseVar = noise_var;
bayes.crossCorr = crossCorr;
// No need to do the initial batch train because we loaded the result
// from .mat.
bayes.WillNeedInitialTrain = false;
return(bayes);
}
public static DNormal FromMatlabStruct(MatlabStruct s)
{
string className = s.GetString("className");
if (!className.Equals(MATLAB_CLASS))
{
throw new ArgumentException("The input does not represent a " + typeof(DNormal));
}
Vector meanVec = s.Get1DVector("mean");
if (meanVec.Count != 1)
{
throw new ArgumentException("mean vector is not 1 dimenion.");
}
double mean = meanVec[0];
double variance = s.GetDouble("variance");
return(new DNormal(mean, variance));
}
// construct a RFGMap from MatlabStruct.
// Matlab objects of class RandFourierGaussMap.
// See RandFourierGaussMap.toStruct()
public static RFGMap FromMatlabStruct(MatlabStruct s)
{
// s.className = class(this);
// s.gwidth2=this.gwidth2;
// s.numFeatures=this.numFeatures;
// s.dim=this.dim;
// s.W=this.W;
// s.B=this.B;
string className = s.GetString("className");
if (!className.Equals("RandFourierGaussMap"))
{
throw new ArgumentException("The input does not represent a " + typeof(RFGMap));
}
double gwidth2 = s.GetDouble("gwidth2");
int numFeatures = s.GetInt("numFeatures");
// int dim = s.GetInt("dim");
Matrix W = s.GetMatrix("W");
if (W.Rows <= 0 || W.Cols <= 0)
{
throw new Exception("Loaded weight matrix has collapsed dimensions");
}
if (numFeatures != W.Cols)
{
// expect W to be dim x numFeatures
throw new ArgumentException("Loaded weight matrix's #cols does not match numFeatures.");
}
Vector B = s.Get1DVector("B");
// construct object
RFGMap map = new RFGMap();
map.GaussWidthSq = gwidth2;
map.WeightMatrix = W;
map.BiasVector = B;
Console.WriteLine("mapMatrix W's size: ({0}, {1})", W.Rows, W.Cols);
Console.WriteLine("bias vector length: {0}", B.Count);
return(map);
}