How to count the number of consecutive negative values in a vector?

Hello! What is the most time-saving method (using the vectorization style or others but not for loops) to count the maximum number of consecutive negative values in a vector?

The problem is that there are thousands of vectors of the same length and if one contains more than 3 consecutive negative values then it will be removed from the data set. For example, vector v1 will be deleted due to [-1 -1 -3] chain, but v2 will remain:

v1 = [11 2 3 -1 -2 1 -1 -1 -3 1 3 1]
v2 = [3 -1 4 1 33 44 -11 -3 4 11 3 -1]

Can it be get done by using array modifying functions only (like 'repmat', 'diff', 'cumsum' etc. not consuming ones like 'find')?

Many Thanks!

[Stephen23]

"The problem is that there are thousands of vectors..."

That is indeed a major problem of your code design.

Once you put all of your data into lots of separate variables like that then you force yourself to write slow, buggy, complex, obfuscated code when you try to access them. Trying to magically access thousands or variable names is what beginners do because they don't realize that they are making their code and data much slower and buggier.

The simpler, neater, and much more efficient way of accessing data is to use indexing.

The two main ways that beginners end up with thousands of numbered variables in their workspace (magically defining them, or using load) can both be trivially altered to use just one array with indexing, so your best option is to change your code design. Read this to know more:

https://www.mathworks.com/matlabcentral/answers/304528-tutorial-why-variables-should-not-be-named-dynamically-eval

[V. Luong]

Yes, the data set is loaded only once into a massive array and the variables v1 v2 are just for clearer explanation. Thank you for your comment!

Answer 1

[Jonathan Chin]

This might work using cumsum to find all the negatives in a group and hist to count the consecutive negatives

   v1 = [11 2 3 -1 -2 1 -1 -1 -3 -1 1 3 1];
   v2 = [3 -1 4 1 33 44 -11 -3 4 11 3 -1];

   x = [0 cumsum(v1>=0)]; %makes all the values positive increment the sum, so all negative numbers will keep the next number the same
   numNeg=max(hist(x)-1); % find number of consecutive negative by looking counting all the similar numbers and subtracting 1 for the initial increment

[V. Luong]

Thank you. That worked so well!

Answer 2

[Image Analyst]

I don't see how any of the other answers are getting rid of the vector. You'd need to use the clear function for that. You can use regionprops(), in the Image Processing Toolbox to get the lengths of all the negative runs of values, then clear to get rid of the vector. So try this with your vectors.

% Find lengths of negative values.
props = regionprops(vec < 0, 'Area');
% If 3 negatives in a row or longer, clear vector - get rid of it completely.
if max([props.Area]) >= 3
  clear vec;
end

I call it vec, but you'd call the code using v1, v2, v3, v4, ...... v9000. You could put it into a function like

function deleteIt= ClearIf3Negatives(vec)
deleteIt = false;
% Find lengths of negative values.
props = regionprops(vec < 0, 'Area');
% If 3 negatives in a row or longer, clear vector - get rid of it completely.
if max([props.Area]) >= 3
  deleteIt = true;
end

Then call it like

if ClearIf3Negatives(v1), clear(v1), end;
if ClearIf3Negatives(v2), clear(v2), end;
...
if ClearIf3Negatives(v9000), clear(v9000), end;

But PLEASE tell me you don't actually have thousands of separately named vectors like you said. That would be such bad programming practice.

[V. Luong]

Thank you very much! Never have I used any functions in Image Processing Toolbox. I will try and see if your solution is faster than Jonathan's.

Answer 3

[Birdman]

Download the file from the link and use rcumsum function.

https://www.mathworks.com/matlabcentral/fileexchange/28685-rcumsumc

 idx=find(rcumsum(v1<0)==3);
  v1([idx-2:1:idx])=[];

(EDITED)

I understood that those consecutive negative values should be deleted from the array, therefore my answer looks like this. To delete v1, just type

idx=find(rcumsum(v1<0)==3);
if ~isempty(idx)
   clear v1;
else
end

[V. Luong]

Many thanks to you! I just saved your function 'rcumsum' for upcoming projects. As we are currently working as a team of dozens of members and our highest priority is stability so we have to use built-in functions only

[Image Analyst]

regionprops(), max(), and clear() (used in my solution) are built in functions. rcumsum() is not a built-in function - it's a user-submitted function, though it's built upon built in functions.

I also question why you need to clear the vector. I virtually never clear variables except if I'm using to many large image arrays that I need to because otherwise the user would run out of RAM memory. Normally, I just let the garbage collector free up the space automatically when the function returns. Why do you think you need to call clear?