Hello.
Le lun. 24 janv. 2022 à 08:21, qiqi tang <[email protected]> a écrit :
>
> *HI Gilles Sadowski,*
>
>
>
> @Overridepublic double[] fit(Collection<WeightedObservedPoint> points) {
> final double[] p = super.fit(points);
> return new double[] {
> constrainedM.value(p[0]),
> constrainedK.value(p[1]),
> p[2]
> };
> }
>
> This method Just maps the end result to the range I want.
You've cut several pieces of the suggested code.
[Also, please keep this discussion within a unique
ML thread.]
Did you try the proposed full set of changes?
Was the output of the code not a local optimum?
The feature you need deserves to be part of the test suite,
so you are welcome to file a report on the issue-tracking
system (see detailed information on the project's page).
You should then rewrite your case in the form of a (JUnit)
unit test (see the "test" part of the source repository).
> I think I need to
> let the return value of the fit method fall within the set range,
What do you mean by
> let the [...] value [...] fall within the [...] range
?
What happens if you don't use the suggested trick?
If the value doesn't settle in that range by itself, it means
that there are local minima.
> rather
> than processing the return value of the fit method twice. Just like example:
>
> @Overridepublic double[] fit(Collection<WeightedObservedPoint> points) {
> return super.fit(points);}
>
> The above method returns a value that is directly in the range of the
> parameter variable I want.
In your previous message, you asked:
> What should I do to limit the parameters 'm' to 1 to 100 and 'k'
> to 100 to 10,000
So indeed, bugs notwithstanding, my proposed solution is
supposed to reach that goal by assuming that the domain
of those parameters are the given ranges.
Regards,
Gilles
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
