If I were implementing https://en.wikipedia.org/wiki/Tabu_search I
would definitely start with a looping implementation. Maybe, after I
had it working, I might see a way of replacing a loop with an
induction, but my intuition suggests that this would likely be
suboptimal.

And, of course, I'd try lifting computations out of the loops, if I
could see how to do that....

Good luck,

-- 
Raul

On Tue, May 26, 2020 at 7:08 AM Anton Wallgren <[email protected]> wrote:
>
> Hi, thank you, missed this mail! You are of course right, more details is 
> better :)
>
> I think that posting complete code for this question might obscure the idea, 
> as its more about a pattern I want to use occasionally. An example would be 
> for a Tabu search algorithm I might know a lower bound for the objective 
> function, so if I during search find a solution whose objective value equals 
> the lower bound, I do not want to continue as I know that I can’t find a 
> better solution. If the lower bound is not tight however, or I do not manage 
> to find such a solution, I wish to abort the search after some set number of 
> iterations, as the search could go on forever.
>
> Regards, Anton Wallgren
> On 26 May 2020, 10:39 +0200, ethiejiesa via Programming 
> <[email protected]>, wrote:
> > Hey Anton,
> >
> > Welcome to the J-verse!
> >
> > FWIW, the question as stated makes me suspect that you're (unconsciously)
> > trying to fit J into an imperative programming paradigm. I would venture 
> > that
> > you'd get better answers by more directly sharing what you're trying to
> > accomplish and the problems thus encountered.
> >
> > One simple formula is to simply share the input you have and output you 
> > want.
> > From lurking here on the forums for a bit, those kinds of questions often 
> > get a
> > lot of good attention. It's often the case that one of the old guard peeps 
> > in
> > with a short and sweet solution that is both beautiful and humbling.
> >
> > Cheers,
> >
> >
> > Anton Wallgren <[email protected]> wrote:
> > > Hi, thanks! But as Raul pointed out, this does not allow me to specify 
> > > max iterations.
> > >
> > > Regards, Anton Wallgren
> > > On 26 May 2020, 08:39 +0200, 'Rob Hodgkinson' via Programming 
> > > <[email protected]>, wrote:
> > > > Hi Anton, welcome to J.
> > > >
> > > > This is a further parameter to the power operator (^:) described here:
> > > >
> > > > https://code.jsoftware.com/wiki/Vocabulary/Loopless 
> > > > <https://code.jsoftware.com/wiki/Vocabulary/Loopless> Section “Types of 
> > > > Loops” and the row in the table “Apply a verb repeatedly”, “Until a 
> > > > condition is met”.
> > > > Use Power ([x] u^:v^:_ y)
> > > >
> > > > For your example, double while a condition (eg let’s say while the 
> > > > sequence is < 100 and stop with the value that breaches that condition 
> > > > …)
> > > >
> > > > 2&* ^:(100>])^:_ (1) NB. Sequence here is 1,2,4,8,16,32,64,128
> > > > 128
> > > >
> > > > 2&* ^:(100>])^:_ (5) NB. Sequence is 5, 10, 20, 40, 80, 160
> > > > 160
> > > >
> > > > Best, Rob
> > > >
> > > > > On 26 May 2020, at 4:09 pm, Anton Wallgren <[email protected]> 
> > > > > wrote:
> > > > >
> > > > > Hello!
> > > > >
> > > > > Fairly recent J enthusiast here. I’m wondering about the idiomatic 
> > > > > way to iterate at most n times? I.e. do f^:n y, but with the 
> > > > > possibility of an early exit if some condition is met. Is it (u F. ]) 
> > > > > y, where u is f but with some Z:’s added? E.g
> > > > >
> > > > > f=: 2&*
> > > > > MAX=: n
> > > > >
> > > > > u=: monad define
> > > > > _2 Z: -.*MAX=: MAX - 1
> > > > > _2 Z: some other condition
> > > > > f y
> > > > > )
> > > > >
> > > > > But then you need to globally assign and reassign MAX and this 
> > > > > doesn’t feel very elegant. Another option of course is to use a 
> > > > > for-loop with break.
> > > > >
> > > > > Thanks, Anton Wallgren
> > > > > ----------------------------------------------------------------------
> > > > > For information about J forums see http://www.jsoftware.com/forums.htm
> > > >
> > > > ----------------------------------------------------------------------
> > > > For information about J forums see http://www.jsoftware.com/forums.htm
> > > ----------------------------------------------------------------------
> > > For information about J forums see http://www.jsoftware.com/forums.htm
> >
> >
> > ----------------------------------------------------------------------
> > For information about J forums see http://www.jsoftware.com/forums.htm
> ----------------------------------------------------------------------
> For information about J forums see http://www.jsoftware.com/forums.htm
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

Reply via email to