Filtering a matrix with different rules for each columns

Beginner here,

With an array AB=100x2, I'm trying to retain rows that satisfy different rules for each column at the same time, i.e. AB(i,1)>0.5 and AB(i,2)<0.2

Based on my limited knowledge I've been able to come up with the following:

 AB=rand(100,2); 
 for i=1:length(AB)    
   if AB(i,1)>0.5 & AB(i,2)<0.2;
       C(i,:)=AB(i,:);
   end
 end

Problem with this is that the result contain zeros in the rows that do not meet this condition. How can I make sure that the resulting array is only as big as the number of rows that satisfy this condition? Or is there a simpler/different way to do this? TIA for your inputs.

Answer 1

[Sean de Wolski]

Vectorized with logical indexing:

C = AB(AB(:,1)>0.5 & AB(:,2)<0.2,:);

[Evan]

Small correction if Thomas is looking to retain all columns:

C = AB(AB(:,1)>0.5 & AB(:,2)<0.2,:);

[Thomas]

Sean and Evan:

Thank you both for the answer. It works wonderfully.

[Sean de Wolski]

Good catch Evan!

[Jason Chan]

Thank you Sean and Thomas! I have been looking for something like this. I was wondering how I can do this when my data is a Table (with names and numbers). For example, I have a table with student names and grades, and I would like to filter the table in the same way.

Thanks in advance.

Jason

Answer 2

[Evan]

Your indexing is causing the zeros. If, for example, rows 1 and 2 meet the criteria but row 3 doesn't, i = 4 on the next iteration. Your matrix C is, at that point, still 2 rows long. If row 4 meets the criterion, your code tries to place AB(4,:)'s values in C(4,:).

But there is, as of yet, no C(3,:). So a row of zeros is created in the same way that entering M(4) = 1 into the command window returns M = [0 0 0 1];

An inelegant but easy way to do it would be to remove [0 0] rows at the end. So just put a line at the end of your script that returns the rows where both values aren't zero:

   M = C(sum(C,2)~=0,:);

You could modify your code to prevent this from happening altogether if you wish. Vectorizing the code could make the entire process possible in a single line (avoiding a for loop altogether) and is probably the best choice (see Sean's answer). However, the above works as a sloppy fix.

[Thomas]

Thank you for the insight. As you said, I don't need the removing zero function but I think the information is definitely useful for me in future. Still have a lot to absorb with this program.

[Evan]

No problem, and best of luck in the future. :)