2

I am trying to perform a comparison between the rows of two matrices A and B with the same number of columns.

In matlab the command ismember(a, b, 'rows') returns a vector containing 1 where the rows of A are also rows of B and 0 otherwise, and also returns the highest index in B for each element in A that is a member of B. 

[tf, index] = ismember(A, B, 'rows');

Is there an equivalent function in python? Any ideas how to do it?

ROMANIA_engineer
  • 54,432
  • 29
  • 203
  • 199
Cristina
  • 21
  • 2

2 Answers2

4

you can get your vector as 

same_rows = [a == b for a,b in zip(A, B)]

Note that this will yield True and False instead of 1 and 0 but bool is subclassed from int and True == 1 and False == 0.

to get the max row where this occurs, you can just use

max_row = next(i for i, row in enumerate(reversed(same_rows)) if row == True)

If you want the number of rows that they have in common, you can just use

same_count == sum(same_rows)

Note that this is all for python and assumes that matrices are lists of lists or tuples or tuples of lists or tuples. HTH.

aaronasterling
  • 68,820
  • 20
  • 127
  • 125
  • or you could just use `map( eq, A, B)` using `eq` from `operator`, also note in Python 3.x that bool is no longer an int but it's own type. – wheaties Sep 29 '10 at 18:52
  • 1
    @wheaties: bool is still an int, in Python 3. In fact, `issubclass(bool, int)` returns True. You can have a look at the full discussion: http://stackoverflow.com/questions/2764017/is-false-0-and-true-1-in-python-an-implementation-detail-or-guaranteed-by-t – Eric O. Lebigot Sep 29 '10 at 19:16
  • Note that same_rows = [a == b for a,b in zip(A, B)], gives me yield only True where the rows are equal only at corresponding rows (e.g a[0,:] = b[0,:]) but i need to find all the positions where a row of a is in b. – Cristina Oct 01 '10 at 13:48
1

The ismember library can be helpful.

pip install ismember

Example:

# Import library
from ismember import ismember

# Example with random matrices
a_vec = np.random.randint(0,10,(5,8))
b_vec = np.random.randint(0,10,(5,10))

# Row-wise comparison
Iloc, idx = ismember(a_vec, b_vec, 'rows')

# These should all be True
for i in np.arange(0,a_vec.shape[0]):
    np.all(a_vec[i,Iloc[i]]==b_vec[i,idx[i]])
erdogant
  • 1,544
  • 14
  • 23