def apft(self, over_sampling):
N = len(self.frequencies)
# choosing the samllest frequence on which
# the timelevels will be discretized
freq_min = np.min(np.absolute(self.frequencies))
timelevels = 6 * np.arange(0, 1., 1. / float(over_sampling)) / freq_min
ap_computation = HbComputation(frequencies=self.frequencies,
timelevels=timelevels)
# computing the corresponding hbt_matrix
ap_idft_matrix = ap_computation.ap_idft_matrix()
for line in range(0, 2 * N + 1):
# gramm-schmidt kinf of orthonormalization
ap_idft_matrix[line + 1:, :] -= np.sum(np.dot(ap_idft_matrix[line + 1:, :],
np.diag(ap_idft_matrix[line, :])),
axis=0) / \
np.sum(ap_idft_matrix[line, :] ** 2) * \
ap_idft_matrix[line, :]
# choossing the vector that was most othogonal to
# the one of line
k = np.argmax([np.linalg.norm(ap_idft_matrix[line_ind],
np.inf)
for line_ind in range(line, len(timelevels))])
# swapping vectors
ap_idft_matrix[[line, k], :] = ap_idft_matrix[[k, line], :]
timelevels[[0, k]] = timelevels[[k, 0]]
del ap_idft_matrix
timelevels = timelevels[:2 * N + 1]
timelevels.sort(0)
return timelevels
def _conditionning(self, timelevels):
hb_computation = HbComputation(frequencies=self.frequencies,
timelevels=timelevels)
return abs(hb_computation.conditionning() - self.target)
def _conditionning(self, timelevels):
ap_computation = HbComputation(frequencies=self.frequencies,
timelevels=timelevels)
return ap_computation.conditionning()
