Sorting arrays in NumPy by column

Question

How can I sort an array in NumPy by the nth column?

For example,

a = array([[9, 2, 3],
           [4, 5, 6],
           [7, 0, 5]])

Answer 1

I had a similar problem.

My Problem:

I want to calculate an SVD and need to sort my eigenvalues in descending order. But I want to keep the mapping between eigenvalues and eigenvectors. My eigenvalues were in the first row and the corresponding eigenvector below it in the same column.

So I want to sort a two-dimensional array column-wise by the first row in descending order.

My Solution

a = a[::, a[0,].argsort()[::-1]]

So how does this work?

a[0,] is just the first row I want to sort by.

Now I use argsort to get the order of indices.

I use [::-1] because I need descending order.

Lastly I use a[::, ...] to get a view with the columns in the right order.