def calculateFirstProbabilies(initialProbabilities, emissionI, emissionD, sequences):
firstCol = [0,0,0,0]
for i in range(4):
initialProb = initialProbabilities[i]
emissionValue = generateEmissionValue(0, i, sequences, emissionI, emissionD)
firstCol[i] = initialProb* emissionValue
return firstCol
def calculateBigADict(postTable, laplaceOffset,sequences, forwardTable, backwardTable, transitionTable, emissionI, emissionD, likelihood):
tableLength = len(postTable)
results = {}
for i in range(4):
for j in range(4):
results[(i,j)] = laplaceOffset
for t in range(0, tableLength-1):
for i in range(4):
for j in range(4):
eValue = generateEmissionValue(t, j, sequences, emissionI, emissionD)
numerator = forwardTable[t][i]
numerator = numerator + math.log(transitionTable[i][j])
numerator = numerator + math.log(eValue)
numerator = numerator + backwardTable[t+1][j]
denominator = likelihood
total = numerator - denominator
prob = math.exp(total)
if prob == 0:
print "error in big A dict"
exit(1)
results[(i,j)] = results[(i,j)] + prob
return results
def calculateForwardAlgoLog(initialProbabilities, transitionProbalities, emissionI, emissionD, sequences ):
results = [[0,0,0,0]]
results[0] = calculateFirstProbabiliesLog(initialProbabilities, emissionI, emissionD, sequences)
endLength = len(sequences[0])
for t in range(1, endLength):
currentCol = [0,0,0,0]
for j in range(4):
singleValues = []
emissionValue = generateEmissionValue(t, j, sequences, emissionI, emissionD)
for i in range(4):
singleValues.append(results[t-1][i] + math.log(transitionProbalities[i][j]) + math.log(emissionValue))
currentCol[j] = sumAllLogProbailities(singleValues)
results.append(currentCol)
return results;
def calculateForwardAlgo(initialProbabilities, transitionProbalities, emissionI, emissionD, sequences ):
results = [[0,0,0,0]]
results[0] = calculateFirstProbabilies(initialProbabilities, emissionI, emissionD, sequences)
endLength = len(sequences[0])
for t in range(1, endLength):
currentCol = [0,0,0,0]
for j in range(4):
singleValue = 0
emissionValue = generateEmissionValue(t, j, sequences, emissionI, emissionD)
for i in range(4):
singleValue = singleValue + (results[t-1][i]*transitionProbalities[i][j])
currentCol[j] = singleValue * emissionValue
results.append(currentCol)
return results;
def calculateBackwardsAlgo(transitionProbalities, emissionI, emissionD, sequences ):
results = [[1,1,1,1]]
endLength = len(sequences[0])
for t in range(1, endLength):
currentCol = [0,0,0,0]
for i in range(4):
singleValue = 0
for j in range(4):
indexInSequence = endLength - t;
emissionValue = generateEmissionValue(indexInSequence, j, sequences, emissionI, emissionD)
singleValue = singleValue + (results[0][j]*transitionProbalities[i][j]*emissionValue)
currentCol[i] = singleValue
results.insert(0,currentCol)
return results;
def calculateBackwardsAlgoLog(transitionProbalities, emissionI, emissionD, sequences ):
results = [[math.log(1),math.log(1),math.log(1),math.log(1)]]
endLength = len(sequences[0])
for t in range(1, endLength):
currentCol = [0,0,0,0]
for i in range(4):
singleValues = []
for j in range(4):
indexInSequence = endLength - t;
emissionValue = generateEmissionValue(indexInSequence, j, sequences, emissionI, emissionD)
singleValues.append((results[0][j]) + math.log(transitionProbalities[i][j]) + math.log(emissionValue))
finalVal = sumAllLogProbailities(singleValues)
if finalVal == float("inf") or finalVal == -float("inf") or finalVal == 0:
print singleValues
exit(0)
currentCol[i] = finalVal
results.insert(0,currentCol)
return results;
def calculateVeterbiEncoding(initialProbabilities, transitionProbalities, emissionI, emissionD, sequences ):
results = [[0,0,0,0]]
results[0] = calculateFirstProbabilies(initialProbabilities, emissionI, emissionD, sequences)
endLength = len(sequences[0])
backPointers = {}
for t in range(1, endLength):
currentCol = [0,0,0,0]
for j in range(4):
singleValue = 0
emissionValue = generateEmissionValue(t, j, sequences, emissionI, emissionD)
backIndex = 0
for i in range(4):
newValue = results[t-1][i]*transitionProbalities[i][j]
if newValue > singleValue:
backIndex = i
singleValue = newValue
backPointers[(t,j)] = backIndex
currentCol[j] = singleValue * emissionValue
results.append(currentCol)
return (results, backPointers);
