Timeline for What would I prefer - an over-fitted model or a less accurate model?
Current License: CC BY-SA 4.0
17 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jul 15, 2021 at 13:48 | answer | added | Pawan Arya | timeline score: 0 | |
| Mar 12, 2021 at 2:19 | answer | added | Nick Corona | timeline score: 1 | |
| Jan 17, 2020 at 19:23 | comment | added | Matthew Drury | As mentioned in @Ray's answer, it is not correct that the first model is necessarily overfit. Random Forests, for example, are explicitly designed to produce this situation. | |
| Jan 13, 2020 at 21:57 | comment | added | David Waterworth | The test and train set metrics are random variables - if for example you're using k-fold cross validation then you can estimate the confidence of both using the variance for example - this could indicate that you cannot really say 1 is higher than 2 with say 90% confidence. So whilst it's a bit hand wavey I would prefer the model which generalises better | |
| Jan 13, 2020 at 21:41 | answer | added | Acccumulation | timeline score: 4 | |
| Jan 13, 2020 at 19:20 | review | Close votes | |||
| Jan 14, 2020 at 4:30 | |||||
| Jan 13, 2020 at 19:17 | answer | added | Ray | timeline score: 9 | |
| Jan 13, 2020 at 11:19 | answer | added | Itamar Mushkin | timeline score: 0 | |
| Jan 13, 2020 at 10:41 | comment | added | Mast | Can we have a third option: use the second to further improve into a third? The second is likely more salvageable than the first, but neither would be ideal in production. What is your goal? | |
| Jan 13, 2020 at 9:27 | comment | added | Will | 1% difference in test accuracy will not be statistically significant for many test sets. How big is yours? | |
| Jan 13, 2020 at 6:55 | comment | added | jerlich | The difference of 1% on the test set makes it easy to choose the 2nd one. But I think if the difference was 10% on the test set than people might choose the "overfit" model. | |
| Jan 13, 2020 at 2:54 | answer | added | Ben Reiniger♦ | timeline score: 16 | |
| Jan 12, 2020 at 21:58 | history | became hot network question | |||
| Jan 12, 2020 at 15:37 | answer | added | FrancoSwiss | timeline score: 4 | |
| Jan 12, 2020 at 14:27 | answer | added | Noah Weber | timeline score: 22 | |
| Jan 12, 2020 at 13:50 | review | First posts | |||
| Jan 12, 2020 at 15:17 | |||||
| Jan 12, 2020 at 13:48 | history | asked | EitanT | CC BY-SA 4.0 |