dive/VDPMLinear.py at master · razvanmarinescu/dive · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132

import voxelDPM
import numpy as np
import scipy
import DisProgBuilder

class VDPMLinearBuilder(DisProgBuilder.DPMBuilder):
  # builds a voxel-wise disease progression model

  def __init__(self, isClust):
    super().__init__(isClust)

  def generate(self, dataIndices, expName, params):
    return VDPMLinear(dataIndices, expName, params)

class VDPMLinear(voxelDPM.VoxelDPM):
  def __init__(self, dataIndices, expName, params):
    super().__init__(dataIndices, expName, params)

  @staticmethod
  def makeThetasIdentif(thetas, shiftTransform):

    #mu, sigma = shiftTransform[0],shiftTransform[1]

    thetas[:,1] += thetas[:,0]*shiftTransform[0] # b = b + a*mu
    thetas[:,0] *= shiftTransform[1] # a = a*sigma

    return thetas

  def initTrajParams(self, crossData, crossDiag, clustProbBC, crossAgeAtScan, subShiftsLong, uniquePartCodeInverse, crossAge1array, extraRangeFactor):
    nrClust = clustProbBC.shape[1]
    thetas = np.zeros((nrClust, 2), float)
    variances = np.zeros(nrClust, float)
    nrSubjCross = crossData.shape[0]
    for c in range(nrClust):
      weightedDataMean = np.average(crossData, axis = 1, weights = clustProbBC[:, c])
      thetas[c,:], residual,_,_ =  np.linalg.lstsq(crossAge1array, weightedDataMean)
      variances[c] = residual/nrSubjCross

    return thetas, variances

  def estimShifts(self, dataOneSubj, thetas, variances, ageOneSubj1array, clustProb,
    prevSubShift, prevSubShiftAvg, fixSpeed):

    objFuncLambda = lambda shift: self.objFunShift(shift, dataOneSubj, thetas,
      variances, ageOneSubj1array, clustProb)
    objFuncDerivLambda = lambda shift: self.objFunShiftDeriv(shift, dataOneSubj,
      thetas, variances, ageOneSubj1array, clustProb)

    res = scipy.optimize.minimize(objFuncLambda, prevSubShift, method='BFGS',
      jac=objFuncDerivLambda, options={'gtol': 1e-8, 'disp': False})

    newShift = res.x
    return newShift

  def objFunShiftDeriv(self, shift, dataOneSubjTB, thetas, variances, ageOneSubj1array, clustProb):

    dps = np.sum(np.multiply(shift, ageOneSubj1array),1)
    #print(ageOneSubj1array.shape)
    nrClust = clustProb.shape[1]
    #for tp in range(dataOneSubj.shape[0]):
    sumSSDalpha = 0
    sumSSDbeta = 0
    for k in range(nrClust):
      errorsSB = dataOneSubjTB - self.trajFunc(dps, thetas[k, :])[:, None]
      betaSqErrorDerivSB = (2 * errorsSB * -thetas[k, 0])
      alphaSqErrorDerivSB = betaSqErrorDerivSB * ageOneSubj1array[:,0][:, None]

      sumSSDalpha += np.sum(np.sum(alphaSqErrorDerivSB, 0) * clustProb[:,k])/ variances[k]
      sumSSDbeta += np.sum(np.sum(betaSqErrorDerivSB, 0) * clustProb[:, k]) / variances[k]

    logPriorShiftDeriv = self.logPriorShiftFuncDeriv(shift, self.paramsPriorShift)

    return np.array([sumSSDalpha - logPriorShiftDeriv[0], sumSSDbeta - logPriorShiftDeriv[1]])

  def estimThetas(self, data, dpsCross, clustProbB, prevTheta, nrSubjLong):

    objFuncLambda = lambda theta: self.objFunTheta(theta, data, dpsCross, clustProbB)
    objFuncDerivLambda = lambda theta: self.objFunThetaDeriv(theta, data, dpsCross, clustProbB)

    res = scipy.optimize.minimize(objFuncLambda, prevTheta, method='BFGS', jac=objFuncDerivLambda,
                                  options={'gtol': 1e-8, 'disp': False})

    newTheta = res.x
    #print(newTheta)
    newVariance = self.estimVariance(data, dpsCross, clustProbB, newTheta, nrSubjLong)

    return newTheta, newVariance

  def objFunThetaDeriv(self, theta, dataSB, dpsCrossS, clustProbB):

    errorsSB = 2*(dataSB - self.trajFunc(dpsCrossS, theta)[:, None])

    errorsSBslope = errorsSB * -dpsCrossS[:, None]
    errorsSBintercept = errorsSB * -1

    meanSSDslope = np.sum(clustProbB[None,:] * errorsSBslope, (0,1))
    meanSSDintercept = np.sum(clustProbB[None, :] * errorsSBintercept, (0, 1))

    return np.array([meanSSDslope, meanSSDintercept])


  def trajFunc(self, s, theta):
    """
    linear function for trectory with params [a,b]
    f(s|theta = [a,b]) = a*s+b

    :param s: the inputs and can be an array of dim N x 1
    :param theta: parameters as np.array([a b])
    :return: values of the linear function at the inputs s
    """

    return theta[0]*s + theta[1]


class VDPMLinearStaticBuilder(DisProgBuilder.DPMBuilder):
  # builds a voxel-wise disease progression model

  def __init__(self):
    pass

  def generate(self, dataIndices, expName, params):
    return VDPMLinearStatic(dataIndices, expName, params)

class VDPMLinearStatic(VDPMLinear):
  def __init__(self, dataIndices, expName, params):
    super().__init__(dataIndices, expName, params)


  def recompResponsib(self, data, crossAge1array, thetas, variances,
                      subShiftsCross, trajFunc, prevClustProbBC):
    return prevClustProbBC