optimalBatch4_repeated.py

import numpy as np
import optimalSampling

class LinearFunction2(optimalSampling.FittingFunctionLSMAP):
    # y = b0+b1*x
    def getPriorMean(self):
        return [None, None, 1]

    def getPriorVar(self):
        return [None, None, 0.01*0.01]

    def getFunction(self, xn, beta):
        return beta[0]+beta[1]*xn+beta[2]*xn*xn

    def getPartialDerivative1(self, xn, yn, beta, i):
        priormu=self.getPriorMean()
        priorvar=self.getPriorVar()
        if i==0:
            return 1
        elif i==1:
            return xn
        elif i==2:
            return xn*xn-(beta[2]-priormu[2])/priorvar[2]

    def getPartialDerivative2(self, xn, yn, beta, i, j):
        priorvar=self.getPriorVar()
        if i==2:
            return -1/priorvar[2]
        else:
            return 0.0

h=LinearFunction2()
h.sigma2=0.1
trueBeta=np.asarray([-0.*0.,-2*0.,1])

X=np.asarray([[0],[0.5],[1]])
y=h.simulateFunctionAtMultiplePoints(X, trueBeta, True)

stepx=0.01

def evaluateDiff(h,y,beta,xrange):
    yp=h.simulateFunctionAtMultiplePoints(xrange, beta, False)
    return np.var(y-yp)

evaluator0=optimalSampling.FIMEvaluator()
evaluator2=optimalSampling.VarEvaluator()

Xrange=[]
for x in np.arange(0,1,stepx):
    Xrange.append([x])
Xrange=np.asarray(Xrange)
ytrue=h.simulateFunctionAtMultiplePoints(Xrange, trueBeta, False)

X,y,beta=optimalSampling.simulateProcess(h,trueBeta,X,y,np.asarray([0,0,0]),30,np.mgrid[0:1+stepx:stepx],evaluator2, verbose=False)

Nreplicates=1000
d=np.zeros((Nreplicates,2))
for n in range(Nreplicates):
    _,_,beta=optimalSampling.simulateProcess(h,trueBeta,X,y,np.asarray([0,0,0]),30,np.mgrid[0:1+stepx:stepx],evaluator0, verbose=False)
    d[n,0]=evaluateDiff(h,ytrue,beta,Xrange)

    _,_,beta=optimalSampling.simulateProcess(h,trueBeta,X,y,np.asarray([0,0,0]),30,np.mgrid[0:1+stepx:stepx],evaluator2, verbose=False)
    d[n,1]=evaluateDiff(h,ytrue,beta,Xrange)
    print("Diff: %d %f %f"%(n, d[n,0],d[n,1]))
np.savetxt("diff02.txt",d)