2

I want to replace the max values of each column of a 2d numpy array with -1: 

b = numpy.array([[1,2,3,4],[5,6,7,8], [9,10,11,12]])
#get the max value of each column
maxposcol = b.argmax(axis = 0)
maxvalcol = b.max(axis = 0)
#replace max values with -1 
for i in numpy.arange(b.shape[1]):
    b[maxposcol[i]][i] = -1

Is there any other way to replace the max values whose positions are given by maxposcol[i]?

If I want to find the n maximal values of each column of my matrix what would you advise me to do? Using a sort? Repeating iteratively the search of max value and replace them at each step?

lizzie
  • 1,506
  • 1
  • 18
  • 31

1 Answers1

3

You can do it like this:

>>> a=np.argmax(b, axis=0)
>>> b[a] = -1
>>> b
array([[ 1,  2,  3,  4],
       [ 5,  6,  7,  8],
       [-1, -1, -1, -1]])
fraxel
  • 34,470
  • 11
  • 98
  • 102
  • Thank you, What do you suggest to find the n maximal values of each column? – lizzie Jun 12 '12 at 06:18
  • @lizzie - [this link should help you ](http://stackoverflow.com/questions/6910641/how-to-get-the-n-maximum-values-in-a-numpy-array), also if you're not aware of: `b[b>6]=-1` it may be of use to you :) – fraxel Jun 12 '12 at 07:21