classifier calculates without finding an end for data sets where it worked before. cefhelper.exe consuming RAM.

Hi everyone,

I´m working on a classification task using the classification learner, after generating a feature table including labels with Matlab. Very basic, just like 27 features and 2 or 3 classes.

I did that a lot of times for several data sets and it always worked until - I don´t know what happened. Are there any reasons, why Matlab might not finish calculating? Any properties I can try changing? I´m sure, that I use the same properties in the Classifier, but is there sth in the background, that influences the process?

I hope my question is not to unspecific, I know it might be hard to find out, without seeing my system, but I thought maybe somebody already had a similar problem. I can provide additional information, if that helps.

I already cleared some of the stuff on the hard disk to have more space. Nothing changed.

Thanks in advance, best regards,

Stef

[prasanth s]

which classifier? is any error or matlab hang or program stuck?

[Stefanie Hörnlein]

I meant the classification learner (sorry), doesn´t matter which model I choose. No errors, not stuck, just running and consuming a lot of RAM.

[Stefanie Hörnlein]

Edit: it looks like not the App uses the RAM but Matlab itself. I have 5 instances of cefhelper running. Any ideas?

[Stefanie Hörnlein]

Just in case anybody reads this who has the slightest idea what I can do:

I uninstalled everything, deleted all files and installed the newest release (2022b).

Still Classification learner lags and there is a huge consumption of RAM by Matlab and several instances of (that changed) MATLABWebUI.exe running (called Java Chromium Embedded Framework (JCEF) Helper in the taskmanager). the program that was running several times and consuming ram before I installed new was called cefhelper.exe.

I´m desperate as I´m working on my PhD and time is running out. Would really appreciate any help.

[Star Strider]

Posting the details of the problem (including the data set or a subset of it, what you are classifying, and what you are tellilng the Classification Learner App to do) could help. It might be possible to define an efficient, specific classifier if we knes more about the problem. I understand that this is your Ph.D. thesis work, so you may not want to discuss specific details. However if we are to help, we need more information.

[Stefanie Hörnlein]

Hi star strider,

thank you for your comment. I added the mat-table that I use for the classification task. It really doesn´t matter which model I choose, it starts to lag right from the beginning, I choose the table, the label and let the first model, that is automatically chosen - Fine tree, train. It´s opening, starting the parallel pool, at least it tells me, it does and then the hourglass is turning forever.

As I said, I don´t think it is in anyway related to my code (that works fine) or the way I use the classification learner, because all that worked fine - even, when I trained all possible models at once or used other data.

It must be something with the cefhelper.exe or java or I don´t know. I found this problem in other conversations but there was never a solution. I don´t really recall what I changed on my PC so I thought maybe someone had the same or a similar problem and fixed it. A similar thing seems to happen with adobe creative cloud (says the internet).

Thanks in advance!

Stef

[Stefanie Hörnlein]

Edit: this is the program consuming my RAM / CPU, it´s located it C:\Program Files\MATLAB\R2021a\bin\win64 - Same with 2022a.

It´s packed to make it possible to upload the file.

Answer 1

[Drew]

As a PhD student, you likely have access to the following through your university:

(1) MATLAB online at matlab.mathworks.com

(2) MATLAB official support.

So, you can run classification experiments using MATLAB online at matlab.mathworks.com, and you can also contact mathworks support about the issue on your machine.

Regarding running experiments on matlab.mathworks.com, I just loaded your classtask_Stef.mat into classificationLearner on matlab.mathworks.com, and the basic default tree model trained very quickly without any issues. The MATLAB version on MATLAB online today was '9.13.0.2145394 (R2022b) Update 3'. A screenshot is shown below. Given there were only 29 observations in the data, I chose leave-one-out cross validation (29 folds) for this small experiment.

[Stefanie Hörnlein]

Hi Drew,

thank you for your extended answer. I already contacted the support before.

I didn´t know, I can use Matlab online, so thank you so much! At first it didn´t work - did the same as before, but now I tried again, I really didn´t change anything, but it worked. Now I cross fingers, it stays like this. -.-

Thanks again,

Stef

[Stefanie Hörnlein]

Update: I fixed it, it was a stupid coding problem. Had nothing to do with the cefhelper or else.