| [stem-dev] STEM Call Notes 9/29/2010 |
|
09/29/2010 Task
and Documentation Tracking On
call: Jamie, Dan, Tom, Judy, Matt, Tero, Stefan, Traci, Werner, Mattias Jamie
will be away next week; Stefan will lead call STEM
Release 1.1.0 ·
Delayed
slightly by server changes at Eclipse ·
Had
no bugs with RC3; none identified on call ·
Will
rename RC3 and post as 1.1.0 ·
Matt
will put item on Newsgroup Server
Side STEM ·
Tero:
Exploring use of CDO for distributed set up ·
All
fields w/ non-generated code have to be changed to getter-setter access ·
Identify
semantics used in EMF would have to be changed where tested for equality or
used in HashMaps/HashSets ·
Hasn’t
noticed performance degradation ·
Discussion ·
Jamie:
CDO requirements for database—mostly extracted? ·
Dan:
CDO feature problem—doesn’t implement feature maps, probably
won’t fix to do so ·
Matt:
Abstracted away from relational databases? ·
Jamie:
Can’t do synchronized parallel where everyone is working on several
relational databases; if goal is to distribute very large graphs, can’t
have every node doing this ·
Matt:
Concerned re commit time ·
Dan:
Need to try out; cost of modifying STEM with other than CDO prohibitive; either
distribute w/ CDO or don’t distribute ·
Jamie:
Do tests w/ CDO first, for single massive computation ·
Tero:
Using CDO support to regenerate Java; hasn’t tried multiple machines ·
Dan:
Problems may not materialize on single machine, but show up w/ multiple ·
Jamie:
Look at how it scales, heat (?) size ·
Stefan:
Exploring parallel STEM w/ Nelder Mead algorithm (from NYU in Matlab) in
automated experiments to fit parameters, varying number of servers, seeing
substantial improvement; will test more ·
Jamie:
Will give value; but doesn’t require partitioning a graph; each machine
running its own graph (not sharing) Bug
of the Week: 261267 (suggestion to close it; deferred from last week) ·
Involved
in partitioning, classifying disease parameters ·
Group
discussion and decision: Close the bug Items
from Participants ·
Jamie:
Wiki documentation on scenarios updated for new release ·
Matt:
Release now on website w/ tag; can now merge back ·
Traci:
Query re initializing when simulating a fictitious flu, moving from sample to
population ·
Jamie:
Working w/ ICDC on flu, used Nedler Mead or steady state values; check
literature values ·
Stefan:
Varying susceptibility a factor in sensitivity ·
Jamie:
Try literature values; will send preprint of paper addressing this to Traci
when submitted ·
Stefan:
Set up and run experiments ·
Traci:
Look at more than infected value ·
Jamie:
Non-IBMers who may want to contribute to STEM, please look at contributor
agreement; Eclipse requires employer’s consent; for non trivial
documentation, also code ·
Werner:
Gave contact info for Jamie to several interested US government employees at
JavaOne ·
Response
from potential Portuguese translators of video? Jamie: No, if problem is the
lack of a toll free number in Brazil, will try to do something ·
Status
of script for short video? Stefan: Working on it ·
Mattias:
Getting his team together; filed legal statement inhouse, hopes to have
completed next week ·
Creating
a tender for programming job for importing formats; what features should
plug-in cover? Jamie: Details of roads, etc.; will talk w/ Mattias off call ·
Status
of graph editor? Jamie: Nothing yet, not a top priority this year; Mattias:
Could use some budget and draft tender, would then take ½ year before it starts ·
Jamie:
Can serve in an advisory role; Matt’s done work on this; graph editor a
bigger job than the importer, would take about 2 weeks’ work; Mattias:
Markus estimates same period of time ·
Jamie:
Working w/ Tom, Stefan, Justin Lessler on mosquito model |