/**
* Compute the JS divergence between an instance and a cluster, used for training data
*
* @param instIdx index of the instance
* @param input statistics of the input data
* @param T the whole partition
* @param t index of the cluster
* @param pi1
* @param pi2
* @return the JS divergence
*/
private double JS(int instIdx, Input input, Partition T, int t, double pi1, double pi2) {
if (Math.min(pi1, pi2) <= 0) {
System.out.format(
"Warning: zero or negative weights in JS calculation! (pi1 %s, pi2 %s)\n", pi1, pi2);
return 0;
}
Instance inst = m_data.instance(instIdx);
double kl1 = 0.0, kl2 = 0.0, tmp = 0.0;
for (int i = 0; i < inst.numValues(); i++) {
tmp = input.Py_x.get(inst.index(i), instIdx);
if (tmp != 0) {
kl1 += tmp * Math.log(tmp / (tmp * pi1 + pi2 * T.Py_t.get(inst.index(i), t)));
}
}
for (int i = 0; i < m_numAttributes; i++) {
if ((tmp = T.Py_t.get(i, t)) != 0) {
kl2 += tmp * Math.log(tmp / (input.Py_x.get(i, instIdx) * pi1 + pi2 * tmp));
}
}
return pi1 * kl1 + pi2 * kl2;
}
/**
* Process the input and compute the statistics of the training data
*
* @return an Input object which holds the statistics about the training data
*/
private Input sIB_ProcessInput() {
double valSum = 0.0;
for (int i = 0; i < m_numInstances; i++) {
valSum = 0.0;
for (int v = 0; v < m_data.instance(i).numValues(); v++) {
valSum += m_data.instance(i).valueSparse(v);
}
if (valSum <= 0) {
if (m_verbose) {
System.out.format("Instance %s sum of value = %s <= 0, removed.\n", i, valSum);
}
m_data.delete(i);
m_numInstances--;
}
}
// get the term-document matrix
Input input = new Input();
input.Py_x = getTransposedNormedMatrix(m_data);
if (m_uniformPrior) {
input.Pyx = input.Py_x.copy();
normalizePrior(m_data);
} else {
input.Pyx = getTransposedMatrix(m_data);
}
input.sumVals = getTotalSum(m_data);
input.Pyx.timesEquals(1 / input.sumVals);
// prior probability of documents, ie. sum the columns from the Pyx matrix
input.Px = new double[m_numInstances];
for (int i = 0; i < m_numInstances; i++) {
for (int j = 0; j < m_numAttributes; j++) {
input.Px[i] += input.Pyx.get(j, i);
}
}
// prior probability of terms, ie. sum the rows from the Pyx matrix
input.Py = new double[m_numAttributes];
for (int i = 0; i < input.Pyx.getRowDimension(); i++) {
for (int j = 0; j < input.Pyx.getColumnDimension(); j++) {
input.Py[i] += input.Pyx.get(i, j);
}
}
MI(input.Pyx, input);
return input;
}
