Machine learning

Hi all,

I need to write a system for optimizing the loading of a truck, based
on a large number of highly intuitive criteria and a good deal of fuzzy
logic. For example, the weight of the products must be evenly
distributed (horizontally and vertically), the fragile things mustn't
be under the heavy ones, one customer's products can't be underneath or
behind another customer's products who is to be delivered afterwards,
the products can sometimes be subdivided and rearranged in a way that
makes them fit better, etc...

There are other factors which make it ever more complex, and the
trouble is that the way it's currently done is based solely on inuition
and experience, so it's very difficult to extract clear rules (because
sometimes the above rules conflict, so it takes intuition to figure out
which way is best). At times, the exact volume and shape of the
products may be unknown, so a good guess has to be made based on the
weight and experience...

So I'm thinking of making a system that allows the user to visually
move things around in 3D (thereby, as a first step, making his task a
little easier), and having it gradually learn the rules of how to place
things (ie, which criteria to give priority to over the others, and
when). I can figure out all the criteria, and I think that over time,
the statistics ought to be able to fine-tune the quality of the spatial
arrangements (the idea being that the application would propose an
arrangement, the user would adjust it if necessary, and the application
would learn from its mistakes). Similar in a way to how many spam are
able to learn what the user considers spam and what he doesn't.

So... my question is: do you have any pointers to guide me in actually
implementing this? Are there any libraries available that would assist
me? I'm unsure whether to use Delphi or Java, or something else, I'm
leaning towards Java but I'm eager to hear your recommendations.

--
Best regards,
Jonathan Neve
_______________
CopyTiger - advanced database replicator for Interbase/Firebird!
Web : http://www.microtec.fr/copycat/ct
_______________________________________
CopyCat - database replication components for Delphi/C++Builder!
Web : http://www.microtec.fr/copycat/cc
.

Relevant Pages

  • Re: Machine learning
    ... on a large number of highly intuitive criteria and a good deal of fuzzy ... arrangements (the idea being that the application would propose an ... able to learn what the user considers spam and what he doesn't. ... I'm unsure whether to use Delphi or Java, or something else, I'm ...
    (borland.public.delphi.thirdpartytools.general)
  • Re: A Comparison of the Influence of the Chinese and Western Philosophies on the Development of TCM
    ... The only reason I quit going was that it was Monday nights, ... On what criteria have you determined that yingyang's post is spam? ... doesn't fit any of the criteria I used when killing spammers' accounts. ...
    (alt.med.fibromyalgia)
  • Re: Belief is the death of intelligence.
    ... only criteria is that a post is not spam because in this case I would ... question if the poster can even read english. ...
    (talk.origins)
  • Re: A comparison of HLA to VC++...
    ... Well, if the criteria cited are important to you, then, yes, there is no ... and HLA is a reasonable choice for them...they may or may not choose HLA... ... language in general, much less HLA, is hardly used at all because of people ... C++ and Java both do ASM ...
    (alt.lang.asm)
  • Re: Criteria for machine readability
    ... like to specify the criteria that such a language is machine readable, ... after all, you are the one claiming that C and Java ... parses them: no; C is pretty difficult to parse, ... If "machine readable" means that the language is close, feature wise, to ...
    (comp.lang.misc)