MATLAB Answers

Loss functions in MATLAB

26 views (last 30 days)
Fatma Elzahraa Ahmed
Fatma Elzahraa Ahmed on 2 Dec 2016
Commented: Jan on 2 Dec 2016
I want to know how to interpret the loss functions results in MATLAB?
On other words, for example if I got 0.3247 as a results of kfoldLoss() function, is this mean that it is a 32.47% or it is a 0.3247%, or how correctly can I define/interpret this obtained result? If we supposed that the output of my question will be 32.47%, then, how can we interpret a result in the help of MATLAB of the kfoldloss() function as 44.9635 where L = kfoldLoss(cvmodel,'mode','individual')?
Thank you very much in advance

  0 Comments

Sign in to comment.

Answers (1)

Jan
Jan on 2 Dec 2016
If the function replies 0.3247, the calculated result is 0.3247. There is no reason to assume, that the value is meant as percent.
The 2nd part of your question is not trivial. As far as I understand you assume that the result of the function should be 0.3247, but Matlab replies 44.9635. Because Matlab is frequently tested, I would dare to assume, that your assumption is wrong. But how could we know this?

  2 Comments

Fatma Elzahraa Ahmed
Fatma Elzahraa Ahmed on 2 Dec 2016
I did not assume the result. what I meant is that for my evaluation, MATLAB output was 0.3247. When I asked this question in other forum they told me that 0.3247 would be equivalent to 32.47%. Therefore, I am just asking if what they told me is right, then how 44.9635, which is a result existed in MATLAB help, would be interpreted?
Note: the cause of my question is because I can obtain the correct rate of my predictor as 0.9825 for example using classperf() which is equivalent to 98.25%. So, I just wanted to clarify how the "loss functions" represent the results? Are they should be interpreted as they are directly or they are a percentages?
Thank you for your reply
Jan
Jan on 2 Dec 2016
The meaning of "%" is 0.01 . 0.32 and 32% are not only equivalent but exactly the same. 44.96 is the same as 4496% .
It is unusual, that a numerical function reply the results in percent. And if kfoldLoss do this, you would find this mentioned in the documentation explicitely.

Sign in to comment.

Sign in to answer this question.