public double sfaError() {
int nPieceWiseParmsExp = expSpikePatternData.getSpikePatternClass().getnPieceWiseParms();
int nPieceWiseParmsModel = patternClassifier.getSpikePatternClass().getnPieceWiseParms();
double possibleWeightForFutureNEAR = nPieceWiseParmsExp - nPieceWiseParmsModel; // coarse!!
double error = 0;
double minError = 0;
double maxMerror = 1;
double maxCerror = 2;
double maxNsfaError = expSpikePatternData.getCurrentDuration() * 0.3;
SolverResultsStat modelStats =
patternClassifier.getSolverResultsStats()[nPieceWiseParmsExp - 1];
int modelNsfaISIs1 = modelStats.getBreakPoint();
int modelNsfaISIs2 = modelSpikePattern.getISIs().length - modelNsfaISIs1;
double[] model =
new double[] {
modelStats.getM1(),
modelStats.getM2(),
modelStats.getC1(),
modelStats.getC2(),
modelNsfaISIs1,
modelNsfaISIs2
};
double[] exp =
new double[] {
expSpikePatternData.getSfaLinearM1().getValue(),
expSpikePatternData.getSfaLinearM2().getValue(),
expSpikePatternData.getSfaLinearb1().getValue(),
expSpikePatternData.getSfaLinearb2().getValue(),
expSpikePatternData.getNSfaISIs1().getValue(),
expSpikePatternData.getNSfaISIs2().getValue(),
};
for (int i = 0; i < 6; i++) {
if (i < 2) { // slopes
error += StatUtil.calculate0to1NormalizedError(exp[i], model[i], minError, maxMerror);
} else {
if (i < 4) { // intercepts
error += StatUtil.calculate0to1NormalizedError(exp[i], model[i], minError, maxCerror);
} else {
error += StatUtil.calculate0to1NormalizedError(exp[i], model[i], minError, maxNsfaError);
}
}
}
error =
patternClassifier.getDynamicFeatWeightMatrix().get(PatternFeatureID.n_sfa_isis2) * error;
if (display) {
displayRoutineForSFA(null, model, error);
}
return error;
}
public double sfa() {
// int nPieceWiseParmsExp = expSpikePatternData.getSpikePatternClass().getnPieceWiseParms();
int nPieceWiseParmsModel = patternClassifier.getSpikePatternClass().getnPieceWiseParms();
SolverResultsStat modelStats = null;
if (nPieceWiseParmsModel > 0) { // ASP, NASP, X
modelStats = patternClassifier.getSolverResultsStats()[nPieceWiseParmsModel - 1];
} else { // ASP error for other comps. for example, TSWB
return 0;
}
return modelStats.getM1();
}
public double sfaErrorObsNormed() {
// System.out.println("****"+sfa()+"****");
// int nPieceWiseParmsExp = expSpikePatternData.getSpikePatternClass().getnPieceWiseParms();
int nPieceWiseParmsModel = patternClassifier.getSpikePatternClass().getnPieceWiseParms();
double error = 0;
SolverResultsStat modelStats = null;
if (nPieceWiseParmsModel > 0) { // ASP, NASP, X
modelStats = patternClassifier.getSolverResultsStats()[nPieceWiseParmsModel - 1];
} else { // ASP error for other comps. for example, TSWB
patternClassifier.calculateAdaptationForNonSP(0);
modelStats = patternClassifier.getSolverResultsStats()[1]; // idx 1 for linear regression
}
int modelNsfaISIs1 = 1 + modelStats.getBreakPoint(); // 0-based idx ; hence +1
int modelNsfaISIs2 = modelSpikePattern.getISIs().length - modelNsfaISIs1; // remaining
double[] model =
new double[] {
modelStats.getM1(),
modelStats.getC1(),
modelNsfaISIs1,
modelStats.getM2(),
modelStats.getC2(),
modelNsfaISIs2
};
PatternFeature[] exp =
new PatternFeature[] {
expSpikePatternData.getSfaLinearM1(),
expSpikePatternData.getSfaLinearb1(),
expSpikePatternData.getNSfaISIs1(),
expSpikePatternData.getSfaLinearM2(),
expSpikePatternData.getSfaLinearb2(),
expSpikePatternData.getNSfaISIs2()
};
// System.out.println(patternClassifier.getDynamicFeatWeightMatrix().size());
float sfa1Weight =
patternClassifier.getDynamicFeatWeightMatrix().get(PatternFeatureID.n_sfa_isis1);
float sfa2weight =
patternClassifier.getDynamicFeatWeightMatrix().get(PatternFeatureID.n_sfa_isis2);
for (int i = 0; i < 6; i++) {
// SFA 1
if (i == 0) { // m1
error += sfa1Weight * NormalizedErrorObsNormed(exp[i], model[i]);
continue;
}
if (i == 1) { // b1
error += sfa1Weight * NormalizedErrorObsNormed(exp[i], model[i]);
continue;
}
if (i == 2) { // nsfaisis1
error += sfa1Weight * NormalizedErrorObsNormed(exp[i], model[i]);
continue;
}
// SFA 2
if (i == 3) { // m2
error += sfa2weight * NormalizedErrorObsNormed(exp[i], model[i]);
continue;
}
if (i == 4) { // b2
error += sfa2weight * NormalizedErrorObsNormed(exp[i], model[i]);
continue;
}
if (i == 5) { // nsfaisis 2
error += sfa2weight * NormalizedErrorObsNormed(exp[i], model[i]);
continue;
}
}
if (display) {
displayRoutineForSFA(exp, model, error);
}
return error;
}
