Find the maximum value in each group of a big matrix

Asked by Yaser Khojah

Yaser Khojah (view profile)

on 26 Apr 2019
Latest activity Commented on by Yaser Khojah

Yaser Khojah (view profile)

on 26 Apr 2019
Accepted Answer by Jan

Jan (view profile)

I have a big matrix and I'm looking for an easy way to group them by a certain way then find the maximum in each group. I have created this code but I got stuck. Anyway to help please. I need to know the indexes of all the maximum points to extract more information from the big matrix.
data = randn(10,1)*10;
edges = 0:5:35';
Y = discretize(data,edges);
[B,I] = sort(Y);
SortedMatrix = [B data(I)];

Yaser Khojah

Yaser Khojah (view profile)

on 26 Apr 2019
so I should have something the below and I want to know ther indices in the original from the original data
1 3.71378812760058
3 11.1735613881447
NaN -0.0684932810334807
Jan

Jan (view profile)

on 26 Apr 2019
So is sortedMatrix your actual input data? Or do you want to process data, which is set to a random array for a demonstration?
It would be easier, if you avoid to create the NaNs, e.g. by adding a final edge Inf. Then splitapply(@max, data(I), B) would solve the problem.
Yaser Khojah

Yaser Khojah (view profile)

on 26 Apr 2019
I have attached my data (Data) and I will explain what I'm trying to do on this file.
edges = 0:5:35';
% i want to split column 18 based on edges
Y = discretize(Data(:,18),edges);
% this is see the new data sorted
[B,I] = sort(Y);
SortedMatrix = [B Data(I,:)];
% How can i find the max in every edges. It might be there is more than a max but I want to know how they are related to the original data so I need their indexs in the original data

on 26 Apr 2019
Edited by Jan

Jan (view profile)

on 26 Apr 2019

data = randn(10,1)*10;
edges = [-Inf, 0:5:35, Inf];
Y = discretize(data, edges);
Result = accumarray(Y, data, [], @max)
I'm confused that this does not work reliably with splitapply:
Result = splitapply(@max, data, Y)
It fails with an error message, if an interval is empty:
For N groups, every integer between 1 and N must occur at least
once in the vector of group numbers.

Rik

Rik (view profile)

on 26 Apr 2019
I was thinking way too complicated here. This was the start of my idea:
nanflag=false;
if any(isnan(B))
replaceNaN=max(B)+1;
B(isnan(B))=replaceNaN;
nanflag=true;
end
b=accumarray(B,Y,[],@max,NaN);
After which removal of empty lines (and putting back the NaN) to get to the desired output should be easy.
Yaser Khojah

Yaser Khojah (view profile)

on 26 Apr 2019
Dear Jan and Rik. Thank you for your help. I'm sorry if i made you both get confused
So, here what I want to do from what I asked earlier. I have this big file (MAT_All). I would like to create this thick line that shows the maximum at every x-axis. So, that is why I was looking for maximum for every interval but not sure if i will get this from what I asked earlier. Any idea?
scatter(MaT_All(:,18), MaT_All(:,17)) 