[FIXED] Efficient way to generate total scores along with numpy.unique()

Issue

My task is as follows:

Each rows of the labels represents students id who received awards from a certain competition. Each competitions have different weighting factors which are given by the weights.

scoresof students are calculated as weighted number of awards that they received. What I have so far is:

import numpy as np

labels = np.array([
    [0,1,5],
    [0,1,3],
    [2,4,5]])
weights = np.array([
    [1],
    [2],
    [4]])

c_labels = np.concatenate(labels)
c_weights = np.concatenate(np.broadcast_to(weights,(3,3)))
uni, inv = np.unique(c_labels,return_inverse=True)

scores = np.zeros(len(uni))
for i in uni:
    scores[i] = np.sum(c_weights[inv==i])
print(scores)

In the above, I have simplified my task as the student scoring problem to explain clearly. This is part of my bigger task, and I need to generate the unique array of the labels and the corresponding scores array to proceed.

The actual data size is huge and I need to reiterate this many times, so I can't use for loops as it takes too long time. I wish I can change to a more efficient way (i.e. replace the last for-loop into an array operation somehow), but I don't know how. Any suggestion would be appreciated.

Solution

Use np.bincount

scores = np.bincount(inv, weights = c_weights)

Answered By - Daniel F