Timeline for My metric is 0.65*accuracy + 0.35*recall. How do I convert that to continuous loss function?
Current License: CC BY-SA 3.0
14 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| S Apr 27, 2017 at 7:46 | history | bounty ended | Hristo Buyukliev | ||
| S Apr 27, 2017 at 7:46 | history | notice removed | Hristo Buyukliev | ||
| Apr 26, 2017 at 8:22 | vote | accept | Hristo Buyukliev | ||
| Apr 25, 2017 at 8:09 | answer | added | Simone | timeline score: 2 | |
| Apr 25, 2017 at 4:59 | answer | added | Matthew Drury | timeline score: 4 | |
| Apr 25, 2017 at 4:37 | comment | added | Matthew Drury | It's not hacky, that's the correct thing to do. Think of it this way, if instead of needing a model, someone told you the true conditional class probabilities $P(y \mid x)$, then you would know everything there is to know about the situation, and your job would necessarily be to set a threshold. | |
| Apr 25, 2017 at 4:07 | answer | added | Meir Maor | timeline score: 2 | |
| Apr 24, 2017 at 21:03 | history | tweeted | twitter.com/StackStats/status/856614701878431744 | ||
| Apr 24, 2017 at 17:11 | comment | added | Simone | If you weight accuracy more than recall it seems to me that you are giving more weight to precision rather than recall. Therefore you do not want false positives. Thus I think the cost weight of a positive sample should be increased. There are ways to make cross-entropy cost sensitive I guess: eg stats.stackexchange.com/questions/68940/… | |
| S Apr 24, 2017 at 17:01 | history | bounty started | Hristo Buyukliev | ||
| S Apr 24, 2017 at 17:01 | history | notice added | Hristo Buyukliev | Draw attention | |
| Apr 24, 2017 at 17:00 | history | edited | Hristo Buyukliev | CC BY-SA 3.0 |
added 58 characters in body
|
| Apr 1, 2017 at 20:29 | answer | added | Hugh Perkins | timeline score: 1 | |
| Apr 1, 2017 at 20:19 | history | asked | Hristo Buyukliev | CC BY-SA 3.0 |