Timeline for Permutation of input features of SVM, simple logistic and KNN classifiers?
Current License: CC BY-SA 4.0
14 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| May 1, 2020 at 19:19 | history | edited | PatternRecognition | CC BY-SA 4.0 |
added 245 characters in body
|
| Oct 29, 2018 at 13:13 | history | edited | Ferdi | CC BY-SA 4.0 |
edited title
|
| Jun 3, 2017 at 14:49 | answer | added | Jan van der Vegt | timeline score: 1 | |
| Nov 4, 2015 at 10:33 | vote | accept | PatternRecognition | ||
| Nov 3, 2015 at 14:56 | answer | added | rep_ho | timeline score: 0 | |
| Nov 3, 2015 at 13:48 | comment | added | PatternRecognition | No, they have got the testing part right, although nothing in this rubbish article actually deserves to be described as "right" . | |
| Nov 3, 2015 at 13:41 | comment | added | PatternRecognition | I think they did so out of desperation for more training data, as there is only a small dataset available for the problem they are attempting to solve. | |
| Nov 3, 2015 at 13:29 | comment | added | rep_ho | Also, since I mentioned double dipping, did they created they test set from the same data the used for feature permutation? I mean could A_1 B_1 C_1 D_1 be in training set and C_1 D_1 A_1 B_1 in the test set? Because that would be major problem, making whole analysis invalid | |
| Nov 3, 2015 at 13:24 | comment | added | rep_ho | I understand. Anyway, I cannot imagine any valid reason for permuting features. also don't consider published articles automatically methodologically correct, it took years until bioinformaticians stopped double dipping with feature selection. | |
| Nov 3, 2015 at 13:13 | comment | added | PatternRecognition | Each one of the 4 parts is represented by 250 features. I used the word "consecutive" only to emphasize the order of the parts. I am sorry, I would rather not sure the paper now, for some reason, but I will add a link to it in a month time when I can. | |
| Nov 3, 2015 at 11:31 | comment | added | rep_ho | I don't get what you mean by 4 consecutive parts and then 1000 features. So you have 10 samples of 1000 features, each measured 4 times in a time serie? Can you just link the paper? I am curious | |
| Nov 3, 2015 at 9:04 | answer | added | Marc Claesen | timeline score: 2 | |
| Nov 3, 2015 at 8:51 | history | edited | PatternRecognition | CC BY-SA 3.0 |
edited title
|
| Nov 2, 2015 at 20:24 | history | asked | PatternRecognition | CC BY-SA 3.0 |