Mistake in the formula for accuracy

[gdalsanto]

flamo/flamo/optimize/surface.py

Lines 305 to 323 in 68d0a71

	def compute_accuracy(self, loss):
	r"""
	Compute the accuracy of the losses according to
	Joseph Turian and Max Henry, “I'm sorry for your loss: Spectrally-based
	audio distances are bad at pitch,” arXiv preprint arXiv:2012.04572, 2020.
	"""
	steps = self.steps
	steps = steps.cpu().numpy()
	# find the index in steps of the element closest to the target value
	target_indx = np.abs(steps - self.param_config.target_value).argmin()
	accuracy = np.empty(loss.shape)
	for i_crit in range(len(self.criteria)):
	for i_run in range(loss.shape[0]):
	for i_step in range(loss.shape[1]):
	accuracy[i_run, i_step, i_crit] = int(
	loss[i_run, i_step, i_crit] > loss[i_run, target_indx, i_crit]
	)
	return accuracy.mean(axis=0)

Does not coincide with Eq. (5) and Eq. (6) of arXiv:2012.04572