a = np.array([10, 20, 30, 40, 50])
ind = np.nonzero(a > 25)[0]
b = a[ind]
creates one more intermediate array and does more lookups so it is slower and less memory efficient compared to "boolean masking":
a = np.array([10, 20, 30, 40, 50])
mask = a > 25
b = a[mask]
Your first approach generates the boolean mask, but then has np.nonzero transform it into an array of indexes, which is not necessary as the boolean mask itself can immediately be used to index the array :)
https://github.com/Igor10798/Internship/blob/9c8a19fcba446cb2e43bf0a8b77ce410f15c4c85/first_network/script.py#L52-L56
Some tips and tricks for
numpy:creates one more intermediate array and does more lookups so it is slower and less memory efficient compared to "boolean masking":
Your first approach generates the boolean mask, but then has
np.nonzerotransform it into an array of indexes, which is not necessary as the boolean mask itself can immediately be used to index the array :)