Talk:Path-based goodness: Difference between revisions
Jump to navigation
Jump to search
No edit summary |
Re |
||
| (2 intermediate revisions by one other user not shown) | |||
| Line 1: | Line 1: | ||
== Choice of parameter values needs justification == | == Choice of parameter values needs justification == | ||
What makes them canonical and what if different values are chosen. Otherwise the data are idiosyncratic and should be noted or kept to the user namespace. [[User:FloraC|FloraC]] ([[User talk:FloraC|talk]]) 19:05, 10 August 2025 (UTC) | What makes them canonical and what if different values are chosen. Otherwise the data are idiosyncratic and should be noted or kept to the user namespace. [[User:FloraC|FloraC]] ([[User talk:FloraC|talk]]) 19:05, 10 August 2025 (UTC) | ||
: I just addressed this. You can also see on [[User:hkm/Sandbox#Badness]] what happens if we increase inaccuracy_fondness up to 0.985--it starts to appreciate less accurate systems more, as one would expect. --[[User:Hkm|hkm]] ([[User talk:Hkm|talk]]) 20:19, 17 August 2025 (UTC) | |||
: I just addressed this. You can also see on [[User:hkm/Sandbox#Badness]] what happens if we increase inaccuracy_fondness up to 0.985--it starts to appreciate less accurate systems more, as one would expect. The overall change is small, especially since we closed the gap between the canonical value and 1 by a full 25%. 72edo is still on top, which is part of why I gave path-based goodness a mention in the 72edo page. --[[User:Hkm|hkm]] ([[User talk:Hkm|talk]]) 20:19, 17 August 2025 (UTC) | |||
:: > The parameters are calibrated according to what best reflects the preferences of composers | |||
:: Of course you should aim for reflecting the preferences of composers. The thing is it adds zero new information about how each parameter translates to that. All the parameters still look like black boxes and I don't see how you can "calibrate" them using any means besides gazing at the end result. | |||
:: [[User:FloraC|FloraC]] ([[User talk:FloraC|talk]]) 06:37, 18 August 2025 (UTC) | |||
Latest revision as of 06:37, 18 August 2025
Choice of parameter values needs justification
What makes them canonical and what if different values are chosen. Otherwise the data are idiosyncratic and should be noted or kept to the user namespace. FloraC (talk) 19:05, 10 August 2025 (UTC)
- I just addressed this. You can also see on User:hkm/Sandbox#Badness what happens if we increase inaccuracy_fondness up to 0.985--it starts to appreciate less accurate systems more, as one would expect. The overall change is small, especially since we closed the gap between the canonical value and 1 by a full 25%. 72edo is still on top, which is part of why I gave path-based goodness a mention in the 72edo page. --hkm (talk) 20:19, 17 August 2025 (UTC)
- > The parameters are calibrated according to what best reflects the preferences of composers
- Of course you should aim for reflecting the preferences of composers. The thing is it adds zero new information about how each parameter translates to that. All the parameters still look like black boxes and I don't see how you can "calibrate" them using any means besides gazing at the end result.