Multipling a multidimensinnal array with a vector

Hi Everyone

I am trying to multiply a multidimensional 13*2560 (13 rows and 2560 coloumns) with 1*2560 (1 row and 2560 coloumns).

But matlab is giving error that matrix dimensions is nit match

For example as for simplest example

A=[1 2 3 4 ; 5 6 7 8]; 2*4 array

>> B=[11 22 33 44 ];1*4 vector

I want first row of A get multiplied with B and give its mean as a output then same for other row.

weighted_mean = mean(B.*A)

Please help

Thanks in advance

Answer 1

[Star Strider]

If you want the mean of each row, specify the dimension argument as 2:

weighted_mean = mean(B.*A, 2);

That should do what you want.

EDIT —

To calculate a mean of the rows with respect to the elements of the weighting vector (as described in Mathematical definition (link) of the Weighted aritmentic mean), this works:

WeightedMean = sum(bsxfun(@times, B, A),2) / sum(B);

equivalently (R2016b and later releases):

WeightedMean = sum(B.*A,2) / sum(B);

This of course assumes that ‘B’ is the weighting vector.

Experiment to get the result you want.

[John Hock]

@ Star Strider

Sir Thanks alot for ur time but it still giving the same error

I dont know what i am missing

I had selected the simplest example but still the same error.

[Star Strider]

I gather you are using MATLAB version R2016a or earlier.

This should work for the true weighted mean, regardless of your MATLAB release:

WeightedMean = sum(bsxfun(@times, B, A),2) / sum(B);

If you only want the unweighted row mean of the product of ‘A’ and ‘B’, this will work, again regardless of your MATLAB version:

RowProductMean = mean(bsxfun(@times, B, A),2);

The R2016b version introduced ‘automatic implicit expansion’, so the bsxfun call would not have been necessary (although it remains highly desirable because the result will be as intended).

[John Hock]

Thanks Sir

Its working

[Star Strider]

As always, my pleasure!

[John Hock]

@Star Strider

Sorry to disturb you again but i again facing some problem

My matrix whose which I want to calculate weighted average z3(22*2562) and weighted vector

abc(1*2562)

Using the above mentioned command

WeightedMean = sum(bsxfun(@times, abc, z3),2) / sum(abc)

But I am expecting the output 1*2562 but it is giving output of just 22*1 values

Pictures are attached

Please help me in this regard

Thanks in advance

[Star Strider]

No worries.

If you want a (1 x 2562) result, sum down the columns rather than across the columns:

WeightedMean = sum(bsxfun(@times, abc, z3),1) / sum(abc);

or using the default column sum:

WeightedMean = sum(bsxfun(@times, abc, z3)) / sum(abc)

My only concern is that may not be the same thing as the true weighted mean, since you are applying the weighting vector across the columns, although summing down the columns.