Implementing an equation in Matlab

Hi,

I am trying to implement an equation in a form like, if a*b<1 then, x=-b/N otherwise x=0, where N= size of a. The equation is simple, but the data I recieve for a and b are differnt formet. a is a 4-d numeric array and b is a categorical array. Would you please suggest me how could I implement this equation?

Sample Input

a= val(:,:,1,1)=0.55

val(:,:,2,1)=0.45

val(:,:,1,2)=0.67

val(:,:,2,2)=0.33

b=

0

1

I have tried to solve it my way like following:

[m,n,o,p]=size(a,4);
x=zeros(m,n,o,p);
c=double(string(categorical(b)));
for i=1:p
    for j=1:o
        mul=c(i)*a(:,:,j,i);
        if(mul<1)
            x(:,:,j,i)=-c(i)/p;
        else
            x(:,:,j,i)=0;
        end
    end
end

Is it the right way to solve the equation?

[darova]

mul=c(i)*a(:,:,j,i);            % it is a 2D matrix
if(mul<1)                       % which of elements you want to compare with 1?

[Saugata Bose]

@darova: thank you very much for showing your interest in this question. I want to compare b with each observation of a. Such as,

c(1)*a(:,:,1,1)=0 * 0.55
c(1)*a(:,:,2,1)=0 * 0.45
c(2)*a(:,:,1,2)=1 * 0.67
c(2)*a(:,:,2,2)=1 * 0.33

Actually, my plan is to take the outputs, 'a' from the softmax layer of a deep learning model and compare it to 'b', the categorical vector.

[darova]

WHat does it mean?

c(1)*a(:,:,1,1)=0 * 0.55

[Saugata Bose]

@darova: As the equation is like if the result of a*b is less than 1 then the value of X would be -b/N.

Inb my problem a must be a 4-d numeric array of shape (1 x 1 x i x j), where i is the category, and j is the observation ID and b must be taken as a categorical vector.

a= val(:,:,1,1)=0.67 (it means, the probability of the first observation stays in category 1 is 0.67)

val(:,:,2,1)=0.33( it means, the probability of the first observation stays in category 2 is 0.33)

However, c(1) refers the calue of first category, in my case, it is 0 where as c(2) refers the value of second category which is 1 here.

[darova]

If mul is a variable (1x1) your above code is OK.

mul=c(i)*a(:,:,j,i);

Also you don't have to use else statement because x matrix is already zeros