Re: Modelling in CS and Physics

["U't Say" <utsay@yahoo.com>]

  Hello anh Aiviet,

  OK. We're now atleast talking about the same topic! :)

  - How to "objectified" a structural problem. Taking C -> C++ translation
as a case study.

  One solution that we can approach is to combine the expert system
concept with the AI learning process (learning as we go) to solve it. It's
likely to be a "software patterns recognition" system, which scans through
the C codes, trying to "recognize" certain patterns, then replace with the
matching C++ implementation from a set of OO patterns. If there is a
section that the system could not recognize, then human intervention is
required to "teach" the system a new pattern. This process is to repeat
until no more to translate, or a percentage (ratio) is reached. We can use
the patterns library from the gang of four to start, and build up more as
the system goes. 

  Undoubtly, the system excels at certain kind, but perform s poorly with
another kind of applications. This is when the expert system come to play.
Each system has ability to recognize only a special set of patterns, and
has its own set of OO patterns to apply. For example, a Network Protocols
Objectifier only able to "objectify" (higher rate of successful
translations) of network protocol applications, and not at the graphical
applications. But this is understandable due to a wide range of
applications. Eventually, a network of these Objectifiers can be connected
to solve a wider range of applications.

   What do you think? I am close or way out on the left field?

   Cheers,
   UtSay
---

--- Aiviet Nguyen <aiviet2@yahoo.com> wrote:
> 
> Hi anh Ut Say,
> --- U't Say <utsay@yahoo.com> wrote:
> >    The book provides some sort of analysis, to help
> > the designer in
> > choosing the patterns. 
> 
>   Yes, they do. However, the reader can choose from a list of patterns
> imposed by the authors. The essense of programming is to model or to
> simulate the reality.
> Few subjective patterns cannot cover the reality. For me this is kind
> of cutting the reality and enforce it into a small box.
> 
> 
> Let's say if an application
> > has a set of
> > requirement criteria and one or more patterns able
> > to fulfill; then the
> > patterns' consequences should be consider to narrow
> > down the selections.
> > If it boils down that more than one can do the job
> > then performance and
> > resource (through profiling) should be applied to
> > select the "correct"
> > one.
> 
>   I see patterns as reusable techniques you can use to save some design
> works, but they are far from being tailored for a given problem.
> 
> 
> >    Agreed. But this applies to any programming
> > languages, not just C ->
> > C++.
> 
>   Yes. That is why I used the word objectification of a structural
> code. This is just an example of what I meant.
>  
> >    IMHO, that there is very little "information" to
> > do this kind of
> > translation successfully. 
> 
>    Yes. Business aalways based on the lack of information :-))
> 
> 
> > The biggest problem still remain in how can we
> > exactly describe/represent the problem as the input,
> > so that we can devise a program to generate the solution as output. 
> > 
> 
>     I see clearly that the structural code alone is not enough. We need
> something additional.
> 
> >    All of these OO models & diagrams (UML, ER...)
> > did come to help, but
> > unfortunately these only address the "output"
> > (implementation) part of the
> > problem and not to depict the problem itself.
> 
>     These tool need to much intelligence and manual work. They are just
> an interface for you to enter your design ideas. You still have to go
> step by step interactively.
> 
> Cheers
> Aiviet
> 
> 
> __________________________________________________
> Do You Yahoo!?
> Bid and sell for free at http://auctions.yahoo.com
> 
> 

__________________________________________________
Do You Yahoo!?
Bid and sell for free at http://auctions.yahoo.com