private static MarkovModel makeMarkovModel(
SimpleAlphabet alphabet,
double[][] transitionMatrix,
double[] startProbabilities,
LinkedHashMap<String, double[]> statesAndCorresponingDirichletParameters,
String modelName)
throws Exception {
int[] advance = {1};
SimpleAlphabet anAlphabet = alphabet;
int numberOfStates = statesAndCorresponingDirichletParameters.size();
ArrayList<String> stateNames = new ArrayList<String>();
ArrayList<double[]> arraysOfDirichletParameters = new ArrayList<double[]>();
for (Map.Entry<String, double[]> me : statesAndCorresponingDirichletParameters.entrySet()) {
double oneDirichletPar[] = me.getValue();
arraysOfDirichletParameters.add(oneDirichletPar);
String oneState = me.getKey();
stateNames.add(oneState);
}
// Distribution initiation
Distribution[] dists = new Distribution[numberOfStates];
EmissionState[] emissionStates = new SimpleEmissionState[numberOfStates];
try {
for (int i = 0; i < numberOfStates; i++) {
dists[i] = DistributionFactory.DEFAULT.createDistribution(anAlphabet);
String oneState = stateNames.get(i);
emissionStates[i] =
new SimpleEmissionState(oneState, Annotation.EMPTY_ANNOTATION, advance, dists[i]);
}
} catch (Exception e) {
throw new Exception("Can't create distributions!");
}
// create HMM
SimpleMarkovModel mm = new SimpleMarkovModel(1, anAlphabet, modelName);
// add states to the model
for (State s : emissionStates) {
try {
mm.addState(s);
} catch (Exception e) {
throw new Exception("Can't add states to model!");
}
}
// create transitions
try {
State magic = mm.magicalState();
for (State i : emissionStates) {
mm.createTransition(magic, i);
mm.createTransition(i, magic);
for (State j : emissionStates) {
mm.createTransition(i, j);
}
}
} catch (Exception e) {
throw new Exception("Can't create transitions!");
}
// set up emission scores
for (Iterator<?> i = anAlphabet.iterator(); i.hasNext(); ) {
AtomicSymbol oneSym = (AtomicSymbol) i.next();
// System.out.println(oneSym.getName());
double[] symbolsInThisSymbolAsArrayOfDoubles = makeArrayOfDoublesFromASymbol(oneSym);
for (int d = 0; d < dists.length; d++) {
double dirichletPar[] = arraysOfDirichletParameters.get(d);
double oneDensity =
DirichletDist.density(dirichletPar, symbolsInThisSymbolAsArrayOfDoubles);
// System.out.print( oneDensity + " ");
oneDensity =
oneDensity
+ 100; // because some of these scores are zero, this one unit is to be able to
// take log! 100
dists[d].setWeight(oneSym, oneDensity);
}
// System.out.println();
}
// set transition scores
Distribution transDist;
// magical to others
transDist = mm.getWeights(mm.magicalState());
for (int i = 0; i < emissionStates.length; i++) {
transDist.setWeight(emissionStates[i], startProbabilities[i]);
}
// from each state to others
for (int i = 0; i < emissionStates.length; i++) {
transDist = mm.getWeights(emissionStates[i]);
transDist.setWeight(mm.magicalState(), 0.0001);
transDist.setWeight(emissionStates[i], transitionMatrix[i][i]);
for (int j = 0; j < emissionStates.length; j++) {
if (i != j) {
transDist.setWeight(emissionStates[j], transitionMatrix[i][j]);
}
}
}
// System.out.println("One HMM set up!");
return mm;
} /*makeMarkovModel*/
public LinkedHashMap<SimpleAlphabet, SimpleSymbolList> getAlphabetAndSimpleSymbolList(
Matrix2D m, Sequence sequence) throws IllegalSymbolException {
LinkedHashMap<SimpleAlphabet, SimpleSymbolList> alphabetAndSymbolList =
new LinkedHashMap<SimpleAlphabet, SimpleSymbolList>();
SimpleAlphabet alphabet = new SimpleAlphabet();
List<AtomicSymbol> listOfSymbols = new ArrayList<AtomicSymbol>();
alphabet.setName("ObservedSequenceAlphabet");
int numberofRows = m.rows();
int seqLength = sequence.length();
if (numberofRows != seqLength) {
System.err.print(
"It was assumed your sequence has a length equal to the number of rows of the matrix, but found a case that is not true!");
}
Symbol gap = sequence.getAlphabet().getGapSymbol();
for (int i = 0; i < numberofRows; i++) {
List<Symbol> oneListOfSymbol = new ArrayList<Symbol>(3);
// red is match, green is flanking and blue is background
double redValue = m.get(i, 0);
double greenValue = m.get(i, 1);
double blueValue = m.get(i, 2);
double onesum = redValue + greenValue + blueValue;
boolean isGap = false;
isGap = (sequence.symbolAt(i + 1) == gap);
if (onesum
== 0) { // sum of this value is supposed to be one, but for some rows it sums up to
// zero, this is to ignore those up to time we found out why these naouthy rows
// sums up to zero!
// continue;
// note these three lines is only a dummy solution for positions where red, green and blue
// are summed up to zero!
redValue = 0.3333;
greenValue = 0.3333;
blueValue = 1 - (redValue + greenValue);
}
// make one triplet symbol from three symbols
Symbol redSymbol = AlphabetManager.createSymbol(Double.toString(redValue));
Symbol greenSymol = AlphabetManager.createSymbol(Double.toString(greenValue));
Symbol blueSymbol = AlphabetManager.createSymbol(Double.toString(blueValue));
oneListOfSymbol.add(redSymbol);
oneListOfSymbol.add(greenSymol);
oneListOfSymbol.add(blueSymbol);
// now create symbol and add it to alphabet
AtomicSymbol oneSym =
(AtomicSymbol)
AlphabetManager.createSymbol(
Annotation.EMPTY_ANNOTATION, oneListOfSymbol, alphabet);
alphabet.addSymbol(oneSym);
listOfSymbols.add(oneSym);
}
SimpleSymbolList ssl = new SimpleSymbolList(alphabet, listOfSymbols);
alphabetAndSymbolList.put(alphabet, ssl);
return alphabetAndSymbolList;
} /*getAlphabetAndSimpleSymbolList*/
