is there a way to call vectors one by one from matrix other than "for" loop?

I have a matrix [a1,a2,a3,a4; b,b,b,b; c,c,c,c], now I want to call each column for a pre-defined function f(a,b,c). There result will be 4x1 vector. I know how to do this with "for" loop, just wondering if there is a way to vectorize the calculation so that don't need to use the loop.

Answer 1

[Matt J]

Not for a general function f(). There are ways to hide the for-loop using arrayfun(), but that's not genuine vectorization.

Depending on the form of f(), however, you may be able to pass the entire matrix to f() and have it return a 4xN result whose columns are computed in a vectorized way. We would have to see the definition of f() to say more.

[hengzh47]

f() is not a closed form function, in this specific case, f = exp(llk_garch(x,y)), llk_garch defined as below:

    function [llk_garch_out] = llk_garch(x,y)
    % x =(w,alpha,beta,lambda,gamma)
    r = 1e-5; %risk free rate

    omega = x(1);
    alpha = x(2);
    beta  = x(3);
    lambda = x(4);
    gamma  = x(5);

        N = length(y);

        h = zeros(N, 1);
        for t = 1:N
            if (t == 1)
                h(t) = omega;
            else
                h(t) = omega + alpha * h(t-1)*((y(t-1)-r)/sqrt(h(t-1))-lambda-gamma)^2 + beta * h(t-1);
            end
        end    
        loglik = -0.5 * (log(h) + ((y-r)./sqrt(h)- lambda).^2);

    llk_garch_out = sum(loglik);
    end

[Matt J]

That function does not look like it produces a 4x1 output. It looks like it returns a scalar. Where are a,b,c?

[Matt J]

In any case, you can certainly do more vectorization within the for-loop, so that matrix-valued y can be handled

      [m,n]=size(y);
      h=zeros(m,n);
      h(1,:)=omega;

      for t=1:m   %loop over rows of y

        last=h(t-1,:);
        h(t,:)=omega + alpha * last.*((y(t-1,:)-r)./sqrt(last)-lambda-gamma).^2 ...
               + beta * last;

      end