[drools-research] Re: research ideas
Adrian Giurca
giurca at tu-cottbus.de
Wed Nov 12 09:54:09 UTC 2008
Mark,
Thank you for the useful hints. We start now accommodating the project
and come back with news.
All the best,
-AdrianG
Mark Proctor wrote:
> AdrianG,
>
> Btw I'm also CC'ing the Drools Research Network on this,
> drools-research at redhat.com <mailto:drools-research at redhat.com>, not
> sure if anyone is actually subscribed yet - heh, but might as well
> start puting that facility to work.
>
> This is a current list of student projects, although we haven't
> mainained it that well - i.e. CEP phase one is near enough done:
> http://www.jboss.org/community/docs/DOC-10745
>
> Davide Sottara is doing some fascinating work on uncertainty with
> drools, one of the interesting outcomes of this so far is he's
> building a hybrid network - that is part discrimination network part
> weighted learning network. He's doing this work for his phd, I don't
> know if he is able to find some aspect of this suitable for a masters
> student... I've put him in the CC, to see if he has any input.
>
> On CEP, according to the power of events book, by prof David Luckham,
> we've done about 1% of his vision :(
> That said what drools will have working for 5.0 is support for the 13
> temporal operators, like before, after etc, and support for sliding
> time windows with automatic life cycle management of events for auto
> retraction - using temporal distance calculations. While this is
> pretty much on par with other "commercial" products, there is still a
> lot more intersting things to do:
>
> Event abstraction
> creating events from another events, but keeping the association for
> the reasoning. I tend to refer to this as composite events, but this
> is a mutli levels beast. On the outside it seems simple, but there are
> large depths to research, especially on the association side. Even if
> we find a simple engineer solution for now, there will no doubt be a
> lot more "theory" to explore that we won't have resources to do.
>
> Causality Tracking
> The ability to determine cause->effect relationships
>
> sequence and hierarchy matching events. This is where we are trying to
> detect specific sequences of events like
> A, B, A but what happens if we have A, A, B, A - do we capture the A,
> B, A or do we capture the A, A, B, A and just ignore the second A.
> Then also as you know composite events (see event abstraction above)
> can result from matching other events, creating a hierarchy, so we
> need to detect these hierarchies. I know there is a lot of demand for
> this, and it's something we will probably research ourselves next
> year, but there is probably more researhc for this at an indepth
> academic level, as ours will be engineering focused.
>
> Verification of events
> Extending our vertifier to handle all the temporal analysis, to find
> warnings, hints, mistakes etc.
>
> Testing
> We need new ways to test these "modelling" environments that are
> capable of doing rules, workflow and event procecssing.
>
> If you have the power of events book, you can pretty much pick any
> advanced feature and we won't have it today, nore many commercial engines.
>
> Not sure if any Rete research is interested to students, as mostly
> this is an engineering effort anyway. We have done partitioning, but
> not parallel patching. There is more research in execution control,
> but this is probably an engineering thing. Maybe distributed Rete,
> where it takes the partitions and distributes them onto other JVMS,
> with fault tolerance?
>
> I'm CC'ing kris and edson. Kris is our flow lead, who you already met
> in Romania, he might have some research ideas related to
> workflow/processes. Edson is our CEP lead, and he can add some other
> interesting CEP research ideas, as well as add more comments on the above.
>
> I've also cc'd Alex and AdrianP, who I both know are very interested
> in temporal reasoning research.
>
> Mark
>
>
>
More information about the drools-research
mailing list