CODEWITHSUNDEEP
pythonnumpy
Abhishek Bhatia
Abhishek Bhatia

Reputation: 9806

Indexing a matrix by a column vector

I have a matrix M of size m x n, and column vector of m x 1. For each of m rows, I need to pickup the index corresponding to the value in the column vector minus 1. Thus, giving me answer m x 1. How can I do this?

zb=a1.a3[np.arange(a1.z3.shape[0]),a1.train_labels-1]

zb.shape
Out[72]: (4000, 4000)

a1.z3.shape
Out[73]: (4000, 26)

a1.train_labels.shape
Out[74]: (4000, 1)

a1.train_labels.head()
Out[75]: 
      22
1618  25
2330   1
1651  17
133   17
2360   5


#my column vector a1.train_labels is shuffled. I don't want to unshuffle it.

Upvotes: 1

Views: 64

Answers (1)

Ami Tavory
Ami Tavory

Reputation: 76297

If your 2d array is M, and indices are a 1d array v, then you can use

M[np.arange(len(v)), v - 1]

For example:

In [14]: M = np.array([[1, 2], [3, 4]])

In [15]: v = np.array([2, 1])

In [16]: M[np.arange(len(v)), v - 1]
Out[16]: array([2, 3])

Upvotes: 1

Related Questions