[seek-kr-sms] Re: Follow-up to chat at NESC workflow meeting
ludaesch at sdsc.edu
Fri Jan 30 09:26:23 PST 2004
Sorry for not having been able to follow up earlier.
I agree with what you identify as the need for a logic and
knowledge-based approach. Hence I'm cc-ing SEEK-KR and GEON-KR folks
I'd like to enourage folks to check out the systems that you mentioned
and then identify a common framework for collaboration.
Note that we will have a SEEK meeting in May at NESC.
Would you be able to attend that? That would be a great way to follow
up on this.
>>>>> "rm" == r muetzelfeldt <r.muetzelfeldt at ed.ac.uk> writes:
rm> Dear Bertram,
rm> I'm glad we made contact at the NESC meeting today - seems like there's a
rm> nice bit of complementarity between SEEK's interest in ontologies for
rm> ecological data and corresponding ontologies for ecosystem modelling.
rm> To recap on my own interests:
rm> I've been interested in the methodology of ecological modelling for some 30
rm> years. About 20 years ago I started collaborating with people in the AI
rm> department here in Edinburgh (Alan Bundy, Dave Robertson, Mike
rm> Uschold) on the Eco project, which sort to formalise the models and the way
rm> we can reason with them: this resulted in a book "Eco-Logic : Logic-Based
rm> Approaches to Ecological Modeling" published by MIT Press. About a year
rm> ago I took early retirement in order to be able to concentrate full-time on my
rm> metamodelling interests.
rm> I still have an interest in knowledge-based aspects of modelling, but some 6
rm> or so years ago decided to first promote a culture of 'declarative modelling
rm> within the ecosystem research community. The vehicle for this is Simile (
rm> http://www.simulistics.com ), a visual modelling environment which uses
rm> System Dynamics (stock-and-flow) notation (like Stella, Modelmaker,
rm> Powersim, Vensim...), but, unlike these other packages, supports object-
rm> based modelling: multiple instances of a class of object, with the possibility
rm> of dynamic creation and destruction of instanves, associations (in the UML
rm> sense) between object classes, etc. (However, it is not 'object-oriented' in
rm> the standard sense: I'm happy to discuss if you want to explore this topic.)
rm> Simile is a nice and powerful package, supporting complex models, both in
rm> terms of the number of equations and the number of object instances. It is
rm> written in Prolog and Tcl/Tk for the GUI, but generates C++ for efficient
rm> model execution. We have now formed a spin-out company - Simulistics
rm> Ltd - (I'm a shareholder but not an employee) though we are keen to be
rm> involved as collaborators in interesting research projects, and quite happy to
rm> discuss possible open-source models.
rm> As far as I am concerned, the main role of Simile is as a 'proof-of-concept
rm> demonstrator' for the virtues of a declarative modelling approach in
rm> ecosystem research. I get deeply depressed when I see models
rm> implemented in procedural programming languages, a mode of thinking
rm> that persists in the great majority of the integrated modelling environments
rm> currently being developed - the large number of them is sufficient evidence
rm> of their inadequacy for providing a common framework for modelling. It
rm> seems quite clear to me that we must abstract out the specification of the
rm> model from one particular form of processing the model (running
rm> simulations) - as is commonplace in most other design contexts
rm> (e.g.architecture, electronic circuit design...), and even in other biological
rm> modelling domains: witness the importance of SBML (Systems Biology
rm> Markup Language, http://www.sbml.org ) in modelling cell processes.
rm> Please have a look at p.17 of
rm> http://gaim.sr.unh.edu/Products/News/Summer2003/rgsum03.pdf for an
rm> outline of my views on this subject: rather stating the obvious, I'm afraid, but
rm> what else can I do? I also have a EU position paper coming out on this
rm> topic, which I can send if you're interested.
rm> The paper that I showed you, on Simile as a possible data flow language for
rm> web services integation, is available at
rm> http://www.decmod.org/documents/webservices/index.html . This is a very
rm> sketchy 'thinking-out-loud' note - not implemented, but I think it would be
rm> useful to explore the degree of common ground with the dataflow concepts
rm> that Simile supports and those of purpose-built dataflow/workflow systems,
rm> as you mentioned. Interestingly, I had a good chat with Eric Simon after
rm> talking to you: some of his Powerpoint slides were virtually identical to the
rm> ones in this note, so the possibility of some form of harmonisation with other
rm> tools must be good.
rm> I am currently working on an XML representation of Simile models, with
rm> XSLT transforms for displaying and processing the models in various ways -
rm> including an an approach based on XSSS (XML StyleSheet Switching) to
rm> allow the same model to be viewed in a web browser through different XSLT
rm> 'filters' (analogous to CSS switching for HTML documents).
rm> I would really love to get back into knowledge-based modelling, one aspect
rm> of which is the seamless integration of data, data analysis and modelling
rm> through common ontologies. It seems that this has a certain resonance
rm> with your interests and involvement in SEEK, and it would be great to
rm> explore this in more depth.
rm> Hope you enjoyed Edinburgh - it's a great city - and looking forward to
rm> meeting up again sometime.
rm> Best wishes,
More information about the Seek-kr-sms