Timeline for Training loss = 0, training accuracy =1, validation and test around 85%
Current License: CC BY-SA 4.0
5 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Apr 5, 2021 at 17:21 | comment | added | CasellaJr | Yes, I know the U-sharp of overfitting. I was not sure about these high levels of accuracy of the models. Thanks | |
| Apr 5, 2021 at 15:05 | comment | added | MetaInformation | The reason I avoided answering the first part is because it seems more like an opinion based question. There's no single rule that we can apply to all cases. I'd argue that second model isn't overfitting as loss was continuously decreasing with epochs, usually you'll see a sharp spike in VL when model starts overfitting. | |
| Apr 5, 2021 at 14:54 | comment | added | CasellaJr | For the first part of the problem do you have any advice? | |
| Apr 5, 2021 at 14:54 | comment | added | CasellaJr | So, you mean that I have to choose the best validation accuracy because is like I am doing exercises at home. I have to choose when I am the best at home, because I don't know the result in a new test. Ok, It is clear. Thank you. | |
| Apr 5, 2021 at 14:14 | history | answered | MetaInformation | CC BY-SA 4.0 |