take sum log of selected elements of matrix

I have 1000 length sequence x=[1 3 5 5 4 1 2 2 1 4 5 3 3 1 2 5 4......] and p matrix below.

          1         2         3         4         5
  1    0.0865    0.5096    0.1538    0.1346    0.1154
  2    0.1070    0.4156    0.1317    0.2263    0.1193
  3    0.0789    0.2105    0.0965    0.3070    0.3070
  4    0.1194    0.0806    0.1290    0.3387    0.3323
  5    0.0965    0.1754    0.0658    0.4474    0.2149

for window size 6 I am selecting 6 data points i.e. 1 3 5 5 4 1 so I want to do

log(P(1,3))+log(p(3,5))+log(p(5,5))+log(p(5,4))+log(p(4,1)).

now I am sliding window by 1 to next i.e. adding next element and skipping previous element of sequence i.e. 3 5 5 4 1 2 & want to do same thing i.e.

log(P(3,5)+log(p(5,5))+log(p(5,4))+log(p(4,1))+log(p(1,2))

and this is upto last of sequence i.e. for 1000 length sequence this operation has 995 times. how to do this.

[John D'Errico]

Is the window size of varying length? Or is it fixed at 6 elements?

You state that for a vector if indices of length 1000, there will be 1995 such operations. Sorry, but as you have explained it, it looks like only 995 operations, NOT 1995.

So you need to be more clear when you ask a question. Otherwise, we are forced to guess.

Answer 1

[Image Analyst]

I think since the elements of x could have virtually any index values, you'll have to do this in a for loop. And I assume you know that P and p are different variables since MATLAB is case sensitive. So, something like

x = [1 3 5 5 4 1 2 2 1 4 5 3 3 1 2 5 4......] 
windowLength = 6;
for windowStart = 1 : length(x)-windowLength+1
    out(windowStart) = log(P(1,3)); % Initialize with the Capital P array element.
    for k = 1 : windowLength - 1
        row = x(k);
        col = x(k+1);
        out(windowStart) = out(windowStart) + log(p(row, col));
    end
end

[Torsten]

I think the OP meant something like

x = [1 3 5 5 4 1 2 2 1 4 5 3 3 1 2 5 4......] 
windowLength = 6;
for windowStart = 1 : length(x) - windowLength + 1
    out(windowStart) = 0.0; 
    for k = windowStart : windowStart + windowLength - 2
        row = x(k);
        col = x(k+1);
        out(windowStart) = out(windowStart) + log(p(row, col));
    end
end

Best wishes

Torsten.