University of Chicago, July 25, 2002
Inter-compatibility of models and software systems
Institute of Adaptive and Neural Computation, Edinburgh University
Much research in computational neuroscience is heavily dependent on software development effort, and much modeling work could usefully build on earlier work undertaken with other software systems. There are therefore two desirable levels of compatibility to be achieved in neuroscience software:
- the ability to run the same model with different software.
- reduce the need for re-implementing models in different environments
- help in validating software by comparing results directly with those from other implementations
- provide an implementation independent way of publishing models (such that they can be run right away, compared to simply reading a, frequently incomplete, description)
- prevent users being locked in to particular packages and restricted to the facilities they provide
- the ability to use components of one software system from within another system.
- reduce duplicate development effort
- improve software quality by letting developers work within their expertise (numerics, graphical user interfaces, databasing etc)
- allow the construction of more flexible models without requiring one software package to cover all the desired features.
Developments such as XML and the Grid (www.gridforum.org) provide technological frameworks for achieving these goals, but there are no magic answers. The NeuroML project provides a toolkit for using XML within independent software packages, but does not claim to be a definitive model description language. Indeed, it may be that the easiest way to achieve point (1) above is not to standardize model descriptions but to agree on a much smaller set of standards for model implementations - ie, to get (1) by first doing (2).
This workshop is aimed at model builders and software builders who wish to discuss solutions to these problems. Brief (20 minute) informal presentations are invited on topics including, but not limited to:
- taxonomies and ontologies for model description
- moving models between simulation systems
- experience integrating modeling software
- technologies for data (model) sharing, eg XML, NeuroML
- integration efforts in related disciplines (see eg sbml.org)
- modeling software currently under development
The organizers will also present a brief introductory overview of various technologies including java and XML to explain some of the jargon and the hype that currently surround them.
Participants wishing to make a presentation should contact Fred Howell in advance, email@example.com