vectorize random walk displacement function

Dear

I would like to code a random walk

As a part of the program the routine call a function that calculate +/-1 in one of the three coordinates for all of the n particles the program follow

This function is as follows:

function rndDesp=RandDesp(n)
        rndV=randi(6,n,1);
        rndDesp=zeros(n,3);
        for j=1:n
            if rndV(j)==1       rndDesp(j,1)=1;
            elseif rndV(j)==2   rndDesp(j,1)=-1;
            elseif rndV(j)==3   rndDesp(j,2)=1;
            elseif rndV(j)==4   rndDesp(j,2)=-1;
            elseif rndV(j)==5   rndDesp(j,3)=1;
            elseif rndV(j)==6   rndDesp(j,3)=-1;
            end
        end
end

The result is a vector rndDesp witch have n rows (one for each particle) and three columns witch have two zeros and one +/-1 obiously in the direction (x,y,z) where the random displacement have to be taken.

The number of n particles could be quite long, so I would like to ask if the is any way to vectorize the for-loop and the if-elseif to speed the calculation.

Thanks in advance

Answer 1

[Walter Roberson]

https://www.mathworks.com/matlabcentral/answers/439633-adding-bias-to-random-walk#comment_660127

Note that particular question involved a biased random walk, and that particular comment did not deal with refusing to go back where you came from. You can look at my Answer there to see a non-vectorized technique for keeping track of "heading", which makes it easier to construct the direction logic. The non-vectorized technique can be vectorized by methods similar to what I showed in the comment I linked to, involving cumsum.