When optimizing a parameter in ASCENT mode, Zorro uses "min" values of other parameters.


It would be better/more productive to use the "default" values - which is normally a "best known estimate", 'reasonably good' value, which is often chosen as the "center of the range'.

And for a parameter already optimized - its "best" value should since be used when training the others in the same TrainCycle.

That's what it does.

Indeed, in v2.33, in TrainCycle=1 Zorro uses default param values (rather than mins), but it wasn't in 2.25 to 2.30.

It certainly was.