Il giorno 20/apr/2011, alle ore 19.19, L. Markowsky ha scritto:
>
>
> Subject: Re: [OctDev] request to register as a developer
> From: [email protected]
> Date: Wed, 20 Apr 2011 12:43:49 +0200
> CC: [email protected]; [email protected];
> [email protected]
> To: [email protected]
>
>
> Il giorno 19/apr/2011, alle ore 22.49, L. Markowsky ha scritto:
>
>
>
> Subject: Re: [OctDev] request to register as a developer
> From: [email protected]
> Date: Sun, 17 Apr 2011 20:59:33 +0200
> CC: [email protected]; [email protected]
> To: [email protected]
>
> Hello,
>
> sorry i misread, you told you had no need to implement fuzzy set theory
> operators, i understood the opposite :). So i don't know, let's see what
> Carlo thinks about it. I think that maybe we can rename our package as "fuzzy
> set theory" and your as "fuzzy inference system" or whatever. Anyway it
> sounds strange to me that you didnt implement the fuzzy set theory operators
> to build something that conceptually is on top of them. How have you
> implemented for example the selection of rules to apply if not by min-maxing
> the conditions? I'm just curious about it :)
>
> Piero
>
> Hello,
>
> I've been puzzled by your question. I implemented a subset of the MATLAB
> Fuzzy Logic Toolbox using only publicly-available documentation
> (www.mathworks.com/help/toolbox/fuzzy/). In order to be MATLAB-compatible,
> the design of my package followed their docs. Does that answer your question?
>
> Thanks,
> L.
>
> Hello,
>
> not properly, it's not a matter of having the same API as Matlab fuzzy
> toolbox rather than how it is implemented the composition under the hood. We
> implemented fuzzy set theory operators as they are the basement semantic for
> fuzzy logic (rules in a rulebase are in disgiunction between each other and
> antecedents of the same rule are in conjunction resulting in a maxmin
> composition computarion) as seen in for example "Fuzzy Logic with engineering
> applications" by Timothy Ross around page 140-150. The description of our
> operators can be find in the same book from page 34 to 41 and from 53 to 62.
>
> As i can see from your code (eval_rules_mamdani.m for example) you
> implemented the min computation of the single membership functions of the
> antecedents in a rule and then aggregating by maxing hte results obtained.
>
> Or approach doesn't mix semantic a syntax in such way, so it is quite
> impossible to modify your system to use our functions at this point, it
> should have been done from the principle, i suppose.
>
> Piero
>
> Hello,
>
> Thanks for taking the time to look at the code.
>
> Min and max are not necessarily used for the "and", "or", "implication", and
> "aggregation" methods, although they are among the choices. There are other
> built-in methods, and the user can specify custom functions as well.
>
> About the design of my functions -- my toolkit is meant to be an open-source
> equivalent of the MATLAB Fuzzy Logic Toolbox, so I wanted to adhere to all of
> the publicly-available documentation. Since that's what I followed as my
> design principle, I had to break the evaluation down as you described. If I
> hadn't, my toolkit wouldn't be (mostly) MATLAB-compatible. In particular, the
> format of the output of evalfis (part of the API of MATLAB's Toolbox)
> dictated the way in which I broke down the evaluation.
>
> Thanks,
> L.
>
>
Hello,
yes that's the point, and that's why they your and our package are not
joinable. You do the composition piece by piece, we do it all in once with
matrix of membership function values (also our package comes with other
built-in t-norms and s-norms and supports custom functions).
So let Carlo or someone of the maintainers decide what to do with our package.
I think that both make sense because of different goals. I stress that probably
renaming our package as fuzzy-set or something like that would be better for
user understanding (we named fl-core because we want to build some more levels
on top of it).
Regards,
Piero
------------------------------------------------------------------------------
Benefiting from Server Virtualization: Beyond Initial Workload
Consolidation -- Increasing the use of server virtualization is a top
priority.Virtualization can reduce costs, simplify management, and improve
application availability and disaster protection. Learn more about boosting
the value of server virtualization. http://p.sf.net/sfu/vmware-sfdev2dev
_______________________________________________
Octave-dev mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/octave-dev
