3
votes

So I know that the numpy argmax retrieves the maximum value along an axis. Thus,

x = np.array([[12,11,10,9],[16,15,14,13],[20,19,18,17]])
print(x)
print(x.sum(axis=1))
print(x.sum(axis=0))

would output,

[[12 11 10  9]
 [16 15 14 13]
 [20 19 18 17]]


[42 58 74]

[48 45 42 39]

This makes sense as the sum along axis 1 (row) is [42 58 74] and axis 0 (column) is [48 45 42 39]. However, i am confused of how argmax work. From my understanding, argmax is supposed to return the max number along the axis. Below is my code and output.

Code: print(np.argmax(x,axis=1)). Output: [0 0 0]

Code: print(np.argmax(x,axis=0)). Output: [2 2 2 2]

Where does 0 and 2 come from? I've deliberately used a set of more complex integer values (9..20) so as to distinguish between the 0 and 2 and the integer values inside the array.

2
may i know why it was downvoted - user859385
what do u mean by index of maximum value - user859385
i googled for it and yet i dont understand. I even tried coding it out to understand how it works. Do u think i will go through the trouble of posting a long post and trying it out without googling? - user859385
np.max returns the maximum values along the respective axis; np.argmax returns 'where' those values occur, the index. - hpaulj

2 Answers

5
votes

np.argmax(x,axis=1) returns the index of maximum of in every row.

axis=1 means "along axis 1", i.e, row.

[[12 11 10  9]    <-- max at index 0
 [16 15 14 13]    <-- max at index 0
 [20 19 18 17]]   <-- max at index 0

Thus its output is [0 0 0].

It's similar for np.argmax(x,axis=0), but now it returns the index of maximum of in every column.

0
votes

Correction: axis=0 refers to rows, not to columns. axis=1 refers to columns, not to rows.

x = np.array([[12,11,10,9],[16,15,14,13],[20,19,18,17]])
  print(x)

[[12 11 10  9]
[16 15 14 13]
[20 19 18 17]]

np.argmax(x, axis=0)
array([2, 2, 2, 2] # third row, index 2 of each of the 4 columns

np.argmax(x, axis=1)
array([0, 0, 0]  # first column, index 0 of each of the three rows.