def _find_rOptimal(self,outputOffsetFactArr,errorArr):
test=errorArr.copy()[1:]
test=np.append(test,errorArr[0])
numValidSolutions=np.sum(~np.isnan(errorArr))
numNanInitial=np.sum(np.isnan(errorArr))
numNanAfter=np.sum(np.isnan(test+errorArr))
valid=True
if numNanAfter-numNanInitial>1:
valid=False
elif numValidSolutions<4:
valid=False
elif numNanInitial>0:
if (np.isnan(errorArr[0])==False and np.isnan(errorArr[-1])==False):
valid=False
if valid==False:
return None
#trim out invalid points
outputOffsetFactArr=outputOffsetFactArr[~np.isnan(errorArr)]
errorArr=errorArr[~np.isnan(errorArr)]
fit=spi.RBFInterpolator(outputOffsetFactArr[:,np.newaxis],errorArr)
outputOffsetFactArrDense=np.linspace(outputOffsetFactArr[0],outputOffsetFactArr[-1],10_000)
errorArrDense=fit(outputOffsetFactArrDense[:,np.newaxis])
rOptimal=outputOffsetFactArrDense[np.argmin(errorArrDense)]
rMinDistFromEdge=np.min(outputOffsetFactArr[1:]-outputOffsetFactArr[:-1])/4
if rOptimal>outputOffsetFactArr[-1]-rMinDistFromEdge or rOptimal<outputOffsetFactArr[0]+rMinDistFromEdge:
# print('Invalid solution, rMin very near edge. ')
return None
return rOptimal
def time_rbf_interpolator(self, neighbors, n_samples, kernel):
interp = interpolate.RBFInterpolator(self.y,
self.d,
neighbors=neighbors,
epsilon=5.0,
kernel=kernel)
interp(self.x)
def __init__(self, coords, values, neighbors=None, smoothing=0, **kwargs):
super().__init__(coords, values)
self.rbf_interpolator = interpolate.RBFInterpolator(self.coords, self.values, neighbors=neighbors,
smoothing=smoothing, **kwargs)