Code in mathematician for check_convergence may not be right

[kwalcock]

There may be an absolute value or two missing in

grid/habitus_ui_interface-main/backend/mathematician.py

Lines 287 to 294 in 8670a1e

	def check_convergence(centroids, last_centroids, j):
	diffs = []
	for i, centroid in enumerate(centroids):
	diffs.append(np.mean(centroid - last_centroids[i]))
	if max(diffs) < 0.00001 or j > 100:
	return True
	else:
	return False

If a centroid and last_centroid are [0, 0, 0] and [10, 10, 10], then centroid - last_centroid is [-10, -10, -10] with a mean of -10, which is certainly less than 0.00001.

If a centroid and last_centroid are [0, 0, 0] and [10, 0, -10], then centroid - last_centroid is [-10, 0, 10] with a mean of 0, which is also less than 0.00001.

It doesn't seem like that fits the definition of convergence.

The code may have to be

diffs.append(np.abs(np.mean(centroid - last_centroids[i])))

for the first issue or even

diffs.append(np.mean(np.abs(centroid - last_centroids[i])))

to take care of them both.

[kwalcock]

@Allegra-Cohen, in case you are only notified when directly addressed.

[Allegra-Cohen]

@kwalcock you're right! I think I threw this function together quite quickly so pretty sure I didn't have a reason to skip abs other than making a mistake. Thanks!