The documentation on np.argmax() says that it
Returns the indices of the maximum values along an axis.
The examples given are straightforward:
In[1]: a = np.arange(6).reshape(2,3)
In[2]: a
Out[2]: array([[0, 1, 2],
[3, 4, 5]])
In[3]: np.argmax(a)
Out[3]: 5
In[4]: np.argmax(a, axis=0)
Out[4]: array([1, 1, 1])
In[5]: np.argmax(a, axis=1)
Out[5]: array([2, 2])
Except in the case of
In[4]: np.argmax(a, axis=0)
Out[4]: array([1, 1, 1])
Since 5 corresponds to a[1][2], why is it returning array([1, 1, 1])?
Also, if I assign
In[6]: b=np.array([[[2,3,4],[4,5,6]],[[3,7,1],[2,5,9]]])
In[7]: b
Out[7]: array([[[2, 3, 4],
[4, 5, 6]],
[[3, 7, 1],
[2, 5, 9]]])
and then ask for the maximum value, why do these two return a different value?
In[8]: b.max()
Out[8]: 9
In[9]: np.argmax(b)
Out[9]: 11
Why does np.argmax() return the integer 11, when that number does not even appear in the array?
Function np.argmax() returns the index of the maximum value, not the value.
In case of array a, each row in array a (you are asking per row, by specifying axis=0) has its maximum at index 1, namely 3, 4, and 5. The three rows are [0, 3], [1, 4], and [2, 5]. In case you asked the argmin(), it would have returned array([0, 0, 0]).
The value of 9 is the element at index 11 of the flattened array b.