labels = (np.arange(num_labels) == labels[:,None]).astype(np.float32) error boolean

[aryopg]

can you explain what does it mean?

[luisgc2116]

It is a way to one-hot-encode the labels. If you break it down, you take the number of labels through a numpy array as such:

Say you have two numpy arrays with the labels and target_vectors.
n_labels = 3
labels_np = np.arange(n_labels)
target_vector = np.array([2,1,0])

Now, target_vector[:,None] returns the array [[2],[0],[1]], which is a column vector. If we
labels_np == target_vector[:,None] return the following boolean array:
[[False False True]
[ True False False]
[False True False]]

Now, the last part, astype(np.float32) simply makes these boolean outputs into integers 0 or 1.