How to broadcast an NumPy array into a section of another NumPy array?

Question

I have these arrays.

 a = np.array([[1,2,3],[3,4,5],[4,5,6]])
 b = np.array([-1, -2, -3])

I want to change the some of elements of a with the elements of b like so:

a = [[1, -1, -1],
     [3, -2, -2],
     [4, -3, -3]]

I tried: a[:,1:]=b but got an exception:

ValueError: could not broadcast input array from shape (3) into shape (3,2)

What is the correct way to broadcast b into a?

@hpaulj Thanks that work. I did try a[:, 1:] = b.T to reshape to (3,1) but that did not work. Why did it not work? Seem that b[:, None] is the way to transpose a row vector to a column vector. — Sun Bear
– Sun Bear, Commented Jul 31, 2020 at 23:42
@SunBear: NumPy's idea of a transpose is n-dimensional, not 2-dimensional. It reverses the order of an array's axes. When there's only one axis, this changes nothing. — user2357112
– user2357112, Commented Aug 1, 2020 at 0:13

user2357112 · Accepted Answer · 2020-07-31 21:56:54Z

3

You're trying to treat b as a column, but the broadcasting rules will try to copy b into the rows of a[:, 1:]. That won't work.

Instead, transpose a and copy b into the rows of the transpose:

a.T[1:] = b

answered Jul 31, 2020 at 21:56

user2357112

286k32 gold badges490 silver badges569 bronze badges

Sign up to request clarification or add additional context in comments.

5 Comments

Ehsan Over a year ago

If the memory-layout of a is such that a.T is a copy, this won't affect a, right?

user2357112 Over a year ago

@Ehsan: a.T cannot be a copy. It's not like reshape. a.T[1:] = b is guaranteed to write into a.

hpaulj Over a year ago

It's the reshape after a transpose that makes a copy. Still I prefer reshaping b so it broadcasts.

Sun Bear Over a year ago

Thanks. Your approach works. Will your approach be more computationally demanding than @hpaulj approach given that the size of a is typically larger than b?

Paul Panzer Over a year ago

@SunBear the transpose (like slicing and most reshaping) is done lazily in numpy, only metadata are modified. The cost does therefore not depend on data size.

Collectives™ on Stack Overflow

How to broadcast an NumPy array into a section of another NumPy array?

1 Answer 1

5 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

5 Comments

Your Answer

Sign up or log in

Post as a guest

Related