From: bernard l. <be...@do...> - 2002-05-26 08:56:42
|
i have been looking at the architecture of the ann api and it is likely t= hat it will be redone a little - there are some redundant looking association= s and other which look inefficient and inflexible. will let you know when this is done. .b -----Original Message----- From: jav...@li... [mailto:jav...@li...]On Behalf O= f bertrand moulard Sent: Saturday, 25 May 2002 8:06 AM To: jav...@li... Subject: Re: [Java AI Library Development] FW: Java AI library progress Hi guys I tried to catch up reading the last emails about Anton and the cbr, and = the framework of the api. I still can't spend lot of time thinking about that= as my last exam is next monday, but I noticed that you were talking about a = gui for the project. Does it make sense to develop a gui for that ? i thought= it was about developping an api to run ai functionalities but not a proper software with it's own gui. Anyway tell me about that, maybe I didn't get everything. I think it could be helpful to chat with one of you next week= to get started and see things more clearly, because right now system programming for uni is badly clouding my vision of the AI project :) Is t= he architecture for the NN done or about ? have a good week end cool geeks ! Ichab. At 15:28 24/05/2002 +1000, you wrote: yeah sure I guess, Let him go for it. Why not get him on this list? alex ----- Original Message ----- From: "bernard leckning" <be...@do...> To: <jav...@li...> Sent: Thursday, May 23, 2002 7:32 PM Subject: RE: [Java AI Library Development] FW: Java AI library progress > yeah it is. > > well i enlarged the scope of this project by calling it the java ai library > because, remember, i used to have the dorja neural network library AND = the > dorja ga library. so i built this project as a framework to hold severa= l > api's. i also figured that there would be cross-cutting aspects between many > potential api's so this framework allows for common aspects to be > centralised and shared. > > from a software design perspective i have been using the java package > framework to seperate concerns and libraries (at a higher level). here'= s the > basics of what we have: > > com.dorja.ailib.config: common configuration for ALL libraries encapsulated > within com.dorja.ailib > com.dorja.ailib.util: cross-cutting aspects that can be shared across a= ll > api's > > com.dorja.ailib.ga: the ga api > com.dorja.ailib.nn: the neural network api > > essentially, to introduce cbr, we simply add: > > com.dorja.ailib.cbr > > and voila! > > i may actually ask him if he wants to do it. although he wants to > concentrate on a gui, i would ask him to do a cbr framework and a gui since > a gui alone would serve no purpose in this project. > > it should be noted that within api's there are certain conventions that > should be adhered to in order to present a coherent set of ai libraries > (more on this later). > > what this package design allows us to do is to give a developer the opt= ion > of bundling on the api's he/she wishes to use. e.g. a develop only want= s to > distribute the ga api then he/she includes: > > com.dorja.ailib.config, com.dorja.ailib.util and com.dorja.ailib.ga. > > cross-api functionality is one of my main goals - mainly, the ability t= o > evolve neural networks using a ga. this is fairly simple with the ga an= d nn > api design that does not require any new packages at the com.dorja.aili= b > level or any changes to the com.dorja.ailib.nn api. i won't go into the > details here though. but this is dependent on the api design and > implementation - the java ai library is merely a framework for bringing= it > all together. > > on a side note, it makes no sense to use a ga for evolving a cbr progra= m > since genetic programming is better suited to the task - the difference= s are > subtle but quite important. i can't see the use of cbr and ann though?!= ?! > care to enlighten me? > > later, > .b > > > > -----Original Message----- > > From: jav...@li... > > [mailto:jav...@li...]On Beha= lf > > Of Alexander Whillas > > Sent: Thursday, 23 May 2002 5:21 PM > > To: jav...@li... > > Subject: Re: [Java AI Library Development] FW: Java AI library progre= ss > > > > > > Interesting idea, > > > > I hadn't thought of making the API cover that much ground, but I gues= s why > > not. I suppose a GA might be useful in developing a decision tree. Ma= ybe > > along the lines of evolving behaviour? I don't know, what would be th= e > > challenge here is the integration of the different components. > > How would the > > GA talk to the CBR or ANN or do you define that as system > > specific if so is > > there some help given? Software design issues mainly. > > > > alex > > > > ----- Original Message ----- > > From: "bernard leckning" <be...@do...> > > To: "javaailibrary-development" > > <jav...@li...> > > Sent: Thursday, May 23, 2002 3:42 PM > > Subject: [Java AI Library Development] FW: Java AI library progress > > > > > > > hi guys, > > > > > > this guy is from zimbabwe and works for a research and development > > company. > > > he showed interest in the project a while ago and i thought i would > > present > > > to you his proposal for joining the project to see how everyone fee= ls > > about > > > it. > > > > > > cheers, > > > .b > > > > > > > -----Original Message----- > > > > From: Anton Lokotsch [mailto:alo...@cs...] > > > > Sent: Tuesday, 21 May 2002 11:50 PM > > > > To: be...@do... > > > > Subject: RE: Java AI library progress > > > > > > > > > > > > hi bernard, > > > > > > > > thanks for the detailed information. i trust everything is going well > > > > with the project. i have an idea for future expansion of this jav= a ai > > > > api; i have done quite a lot of work with cbr (case based reasoning) > > > > systems and feel that this would be an excellent machine learning > > > > technique to include in the api later on. the cbr process itself= is > > > > fairly simple and would greatly enhance the broader appeal of the java > > > > ai api. i am currently working on a cbr project, more specificall= y > > > > designing a java frontend for a cbr application which was > > designed using > > > > one of the cbr shells avialable on the market. i would be willin= g to > > > > contribute ideas/time and my somewhat limited coding skills to th= is if > > > > you think it is a good idea. i dont know what funding you > > have currently > > > > for this project, but i am sure we can try and apply through the many > > > > official channels in an attempt to get some > > academic/scientific/research > > > > funding to further this project as it certainly is a > > brilliant idea with > > > > great development potential. > > > > > > > > i look forward to hearing from you, > > > > anton > > > > > > > > ________________________________________________________ > > > > Anton Lokotsch > > > > Software Development Consultant > > > > DSS - Decision Support Sevices Program > > > > iComtek - Information, Communication and Space Technology Divisio= n > > > > CSIR - Council for Scientific and Industrial Research > > > > ________________________________________________________ > > > > Meiring Naude Rd > > > > POBOX 385 > > > > Pretoria > > > > South Africa > > > > 0001 > > > > ________________________________________________________ > > > > www.csir.co.za/icomtek > > > > Tel +27-12-841-2889 > > > > Fax +27-12-841-4720 > > > > ________________________________________________________ > > > > > > > > >>> "bernard leckning" <be...@do...> 04/15/02 03:15PM >>> > > > > hi anton, > > > > > > > > thanks for your interest in what i am doing. i have only recently > > > > enlisted > > > > the help of a friend so there isn't an email list as yet. in any case > > > > there > > > > needs to be more doco (i might set up a wiki in the next week or = 2). > > > > > > > > i am trying to create an api that allows a developer to easily ta= p in > > > > to ai > > > > capabilities with only a basic understanding of the ai paradigm he/she > > > > is > > > > using. > > > > > > > > the two paradigms being coded at the moment is a genetic algorith= ms > > > > api > > > > (based on the design of matt wall's c++ galib - > > > > http://lanceit.mit.edu/ga) > > > > and an artificial neural networks api. > > > > > > > > this means with a few calls to the api you can set up a ga or an = ann > > > > simulation. the ga code is getting pretty complicated with all th= e > > > > types of > > > > genomes that can be implemented and the ann code will be difficul= t too > > > > as i > > > > am trying to cater to an digraph structured ann as well as > > > > self-organising > > > > maps. > > > > > > > > the best place at the moment to get some details is at > > > > http://javaailibrary.sourceforge.net. i have written some thought= s > > > > about it. > > > > you should then try the javadoc on that page or just look the cod= e > > > > yourself > > > > from the cvs repository - although this is in its very early and rough > > > > stages. > > > > > > > > perhaps after i get a wiki up or some other design doco i want to > > > > really get > > > > some ppl's advice and coding skills. as it will expand from ga's = and > > > > ann's > > > > to evolutionary re-inforcement learning (erl) where i will be evolving > > > > neural networks. > > > > > > > > please keep an eye out on our page as i hope to have things ready= to > > > > get > > > > other ppl on board. any comments you have are most welcome. > > > > > > > > .bernard > > > > > > > > > -----Original Message----- > > > > > From: Anton [mailto:alo...@cs...] > > > > > Sent: Monday, April 15, 2002 08:42 > > > > > To: be...@do... > > > > > Subject: Java AI library progress > > > > > > > > > > > > > > > Hi there, > > > > > > > > > > I am very interested in what you guys are doing. Please, I > > > > > would very much like to hear more about what you are doing/ > > > > > planning to do. Do you have a mailing list or something? > > > > > > > > > > Regards, > > > > > Anton > > > > > > > > > > > > > > > > > > _______________________________________________________________ > > > > > > Don't miss the 2002 Sprint PCS Application Developer's Conference > > > August 25-28 in Las Vegas -- http://devcon.sprintpcs.com/adp/index.= cfm > > > > > > _______________________________________________ > > > JavaAILibrary-Development mailing list > > > Jav...@li... > > > https://lists.sourceforge.net/lists/listinfo/javaailibrary-developm= ent > > > > > > > > > _______________________________________________________________ > > > > Don't miss the 2002 Sprint PCS Application Developer's Conference > > August 25-28 in Las Vegas -- http://devcon.sprintpcs.com/adp/index.cf= m > > > > _______________________________________________ > > JavaAILibrary-Development mailing list > > Jav...@li... > > https://lists.sourceforge.net/lists/listinfo/javaailibrary-developmen= t > > > > > _______________________________________________________________ > > Don't miss the 2002 Sprint PCS Application Developer's Conference > August 25-28 in Las Vegas -- http://devcon.sprintpcs.com/adp/index.cfm > > _______________________________________________ > JavaAILibrary-Development mailing list > Jav...@li... > https://lists.sourceforge.net/lists/listinfo/javaailibrary-development > _______________________________________________________________ Don't miss the 2002 Sprint PCS Application Developer's Conference August 25-28 in Las Vegas -- http://devcon.sprintpcs.com/adp/index.cfm _______________________________________________ JavaAILibrary-Development mailing list Jav...@li... https://lists.sourceforge.net/lists/listinfo/javaailibrary-development Bertrand Moulard D=E9veloppement Java +33 5 53 24 75 34 |