Re: [math] redesigning the optimization/estimation packages (phase 2)

[Ted Dunning <te...@gmail.com>]

On Sat, Feb 28, 2009 at 6:52 AM, Phil Steitz <ph...@gmail.com> wrote:

> What is going on here is that the special case - linear regression analysis
> - is being discussed as an example of an optimization problem.
>

Yes.  It is a special case.

But it was germane to the topic, particularly the imposition of a particular
way of converting a residual vector to a quantity to be optimized.  Mostly I
was worried that the API structure for optimization would force itself on
the higher level API's.


> ...  Patches welcome!
>

Point taken.  I apologize for not providing more them.

-- 
Ted Dunning, CTO
DeepDyve

Re: [math] redesigning the optimization/estimation packages (phase 2)

[Luc Maisonobe <Lu...@free.fr>]

Ted Dunning a écrit :
> On Sat, Feb 28, 2009 at 6:52 AM, Phil Steitz <ph...@gmail.com> wrote:
> 
>> What is going on here is that the special case - linear regression analysis
>> - is being discussed as an example of an optimization problem.
>>
> 
> Yes.  It is a special case.
> 
> But it was germane to the topic, particularly the imposition of a particular
> way of converting a residual vector to a quantity to be optimized.  Mostly I
> was worried that the API structure for optimization would force itself on
> the higher level API's.

The LeastSquaresConverter that started the discussion is only a small
helper class. It is not the basis of the general API and can be removed
if it is too awkward.

Please take a look at what I have just checked in. I have improved
slightly the general interfaces (mainly ConvergenceChecker and
Optimizer). I have also adapted DirectSearchOptimizer, NelderMead and
MultiDirectional to implement the new API. The multi-start feature that
was previously only available in direct search optimizers has been
extracted and is now a wrapper for other optimizers, regardless of their
types.

The optimization.general package (former estimation) is the next task.

Luc

> 
> 
>> ...  Patches welcome!
>>
> 
> Point taken.  I apologize for not providing more them.
> 


---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
For additional commands, e-mail: dev-help@commons.apache.org