Timeline for No overfitting but bad prediction
Current License: CC BY-SA 4.0
11 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Mar 8, 2022 at 18:57 | vote | accept | Zehra N. | ||
| Mar 7, 2022 at 7:13 | comment | added | Zehra N. | @DikranMarsupial thank you so much!! | |
| Mar 7, 2022 at 7:13 | comment | added | Zehra N. | @Paul question is so clear. test prediction results are not nice. But my model is not bad. ? | |
| Mar 7, 2022 at 7:12 | comment | added | Zehra N. | @ChristianHennig you can see the Val acc and training acc. | |
| Mar 6, 2022 at 23:52 | comment | added | Dikran Marsupial | The problem is that the training set is approximately balanced, but your test and validation sets have significantly fewer normal than anormal examples, which is why the classifier is presumably over-predicting "normal" in the test set (difficult to tell because of the labelling). Your training set should be representative of the statistical distribution in the test set (and in operation), and that includes the relative frequencies of the classes. You might want to use stratified resampling to form the training, test and validation sets, so they all have the same label distributions. | |
| Mar 6, 2022 at 22:27 | comment | added | Paul | @ZehraN. please be more specific about the part you did not understand. We are not being paid to help you. It is frustrating to hear "I didn't understand" without any explanation and makes people less inclined to help you. | |
| Mar 6, 2022 at 21:56 | comment | added | Christian Hennig | On what basis did you state "no overfitting" in the title? | |
| Mar 6, 2022 at 17:07 | answer | added | stvhuang | timeline score: 3 | |
| Mar 6, 2022 at 15:54 | comment | added | Zehra N. | 1. I didn't understand your first comment, 2. Do you mean the size of the test is different? | |
| Mar 6, 2022 at 14:29 | comment | added | Paul | A couple possibilities - (1) you have been using validation extensively to optimize hyperparameters, and thus it is effectively just part of your training set; (2) test data just looks different from training/validation | |
| Mar 6, 2022 at 13:40 | history | asked | Zehra N. | CC BY-SA 4.0 |