Stuck in a While loop...sometimes

Hi

So the code below works when the conditional values for the second while loop are 5, 10, and 20. However, when I change any of those values (20 --> 81.5 for example), this causes the LAST while loop to freeze, and I must ctrl+c to exit. Why will this code fail when I change some conditional values?

x = rand(1,500);  %Random People Coordinates

y = rand(1,500);

x = 170*x'-85;

y = 170*y'-85;

prev_dist = zeros(500,1);

i=1;

    while i<501    

   prev_dist(i) = (x(i)^2) + (y(i)^2);
      i = i+1;
  end

Dead = 20.5;
Immobile = 54.5;
Slow = 81.5;
living = zeros(500,1);
i = 1;
while i<501
    if prev_dist(i) < (5)^2
        living(i) = 0;
    else 
        living(i) = 1;
    end
    if prev_dist(i) < (10)^2 && prev_dist(i) > (5)^2
        living(i) = 0.5;
    end
    if prev_dist(i) < (20)^2 && prev_dist(i) > (10)^2
        living(i) = 0.8;
    end
    i = i+1;
end

new_dist = zeros(500,1);
xdum = zeros(500,1);
ydum = zeros(500,1);
i=1;
while i<501
    if living(i) == 1 || living(i) == 0.8
        while new_dist(i) < prev_dist(i)
            xdir = rand();
            ydir = rand();
            if xdir < 0.5 %Left
                xdum(i) = x(i)-2;
            elseif xdir > 0.5 %Right 
                xdum(i) = x(i)+2;
            end
            if ydir < 0.5  %Down
                ydum(i) = y(i)-2;
            elseif ydir > 0.5  %Up
                ydum(i) = y(i)+2;
            end
            if living(i) == 1
                new_dist(i) = (xdum(i)^2) + (ydum(i)^2);
            elseif living(i) == 0.8
                new_dist(i) = (0.8)*((xdum(i)^2) + (ydum(i)^2));
            end
        end
    else
        new_dist(i) = prev_dist(i);
    end
    i = i+1;
end

i = 1;
while i<501
    if living(i) == 1 || living(i) == 0.8
        prev_dist(i) = new_dist(i);
    end
    i = i + 1;
end

prev_dist

[Image Analyst]

You have 5 while loops and they are all "while i<501" except for the 4th one which is "while new_dist(i) < prev_dist(i)" so in which condition are you changing a 5, 10, or 20 to 81.5?

Answer 1

[Fangjun Jiang]

I think it is this while loop that causes the problem.

while new_dist(i) < prev_dist(i)

It might be an infinite loop.

All your other while-loop can be changed to for-loop. You really should look into it. For fixed number of iterations, for-loop is a better option. You really don't need all the i=i+1; lines.

[Amy]

But this loop works fully for the way it is written above. The problem occurs for that while loop ONLY when I change the conditions in the second while loop to be my desired values of 20.5, 54.5 and 81.5, instead of 5, 10, and 20, respectively.

Why would this change cause the last while loop to fail?

[Image Analyst]

Like I said in my answer, you should be able to figure this out on your own. Why don't you plot new_dist and prev_dist at the end of the loop and see what they look like?

[Fangjun Jiang]

When you change those parameters, the value of prev_dist is affected.

[Dr. Seis]

No Fangjun, "prev_dist" doesn't change based on the numbers she describes. "prev_dist" is defined before "Dead", "Immobile", and "Slow" so that doesn't explain it. The only place "prev_dist" changes is at the end, which isn't apart/inside of that problematic while loop.

In fact, changing those numbers doesn't appear to affect anything in her code. I wish we knew how those numbers factor in.

[Fangjun Jiang]

Good point, Elige! I assumed too soon.

I found out that even without changing the value from 20 to 81.5, it could also be trapped in the infinite while loop. I think it is data dependent because the code used lots of random number. 81.5 just makes it happen more often.

Without any comments, it's hard to understand the intent of the code. But with enough time, I wouldn't think this is a hard problem to solve.

Answer 2

[Image Analyst]

Normal debugging can solve this. For some reason your " while new_dist(i) < prev_dist(i)" loop never exits when i = 112. I don't know why but I'm sure you can track it down using regular debugging techniques like everyone knows (setting breakpoints, examining variables, etc.) just as well as I can.

[Amy]

Is this for the new values of 20.5, 54.5 and 81.5? And I'm not really familiar with breakpoints adn the actual Matlab debugger. Hopefully I can figure it out.

Answer 3

[Dr. Seis]

The problem is with the population where "living" is equal to (0.8). There are cases where no combination of adding or subtracting 2 from the x- and y-coordinates (then of course squaring the sum/differences, then summing the squares, and then multiplying the sum of the squares by 0.8) will allow the "new_dist" to ever be greater than "prev_dist".

In the following code, I show you visually the whereabouts of the problematic x- and y-coordinate pairs:

[X,Y] = meshgrid(-85:0.1:85);
prev_dist = X.^2 + Y.^2;
prev_dist(prev_dist > (20)^2) = 0;
prev_dist(prev_dist < (10)^2) = 0;

range = zeros(size(X));
range(prev_dist > 0) = 1;

counter = 0;
for ii = 1 : size(X,1)
  for jj = 1 : size(X,2)
      if range(ii,jj)
          x = X(ii,jj);
          y = Y(ii,jj);
          for i = 1 : 4
              for j = 1 : 4
                  r = ((x+2*(-1)^i)^2 + (y+2*(-1)^j)^2)*0.8;
                  if r > prev_dist(ii,jj)
                      counter = counter + 1;
                      break;
                  end
              end
              if r > prev_dist(ii,jj)
                  break;
              end
          end
          if r <= prev_dist(ii,jj)
              range(ii,jj)=0.5;
          end
      end
  end
end

figure;
imagesc(-85:0.1:85,-85:0.1:85,range);
colorbar;
xlabel('x');
ylabel('y');
title(sprintf('Location of data with "living" equal 0.8 has amplitude of 1 or 0.5'));

The x- and y- points that are associated with a "living" value of (0.8) have an amplitude greater than 0. Those points that are able to exit out of the while loop (eventually) have an amplitude of 1. The points that will never, ever, ever be able to exit out of the while loop (under the conditions you set above) will have an amplitude of 0.5.