CODEWITHSUNDEEP
imagematlabimage-processing
Maxxx
Maxxx

Reputation: 3768

Lexicographical sorting of matrix

I'm required to sort the matrix(each row is a feature vector) in lexicographical order. 

test_mat = [10 40 30 -1 ; 43 78 -5 1; 87 54 -4 -2];

But after looking at several posts, there are:

sortrows(test_mat)
sort(test_mat)

I'm not entirely sure which method is right, as I'm getting quite different results from each.

Upvotes: 1

Views: 1459

Answers (1)

avermaet
avermaet

Reputation: 1593

According to Sardar Usama's comment I'm using the test matrix:

A = [10 40 30 -1; 
    43 78 -5 1; 
    87 54 -4 -2];

sort(A,dim) sorts every vector of A in ascending order. If dim is 1, then every column vector, if dim is 2 then ever row vector.

E.g.:sort(A, 2)sorts every row of A in ascending order. This means that every row will be considered a row vector and will be sorted in ascending order.

% ascending sorting of every row vector
sort(A,2)
ans =
    -1    10    30    40
    -5     1    43    78
    -4    -2    54    87

% ascending sorting of every column vector
sort(A,1)
ans =
    10    40    -5    -2
    43    54    -4    -1
    87    78    30     1

sortrows(A, col) sorts the rows of A in ascending order according to the column col. The individual rows don't change, so the rows are not modified.

E.g.:sortrows(A,1)won't change A at all, since the first column is already in ascending order.

sortrows(A,1)
ans = 10    40    30    -1
      43    78    -5     1
      87    54    -4    -2

If you just use sortrows(A), A will always be sorted according to the first column by default, if you use sortrows(A,1), it does therefore the same thing, but you could also use sortrows(A,3) and sort according to the 3rd column. Additionally you can also use sortrows(A,[1 3]), which sorts first based on column 1 and all rows where column 1 has the same value will be sorted according to column 3. sortrows(A,[1 2 3 4]) is of course the same as sortrows(A). (thats the default)

So if your rows are feature vectors, sortrows will order your feature vectors according to one of the features/columns, but the feature vectors/rows itself will stay unchanged and sort will also change the rows/feature vectors itself, since every row will be changed/sorted in ascending order. So I guess you want sortrows.

If you only have numeric values, I don't see any reason for casting.

Upvotes: 4

Related Questions