DS6 Stage01 Preliminary Meeting

Royal Observatory Edinburgh - 25 January 2005

Purpose

The purpose of this meeting was to discuss general ideas for DS6 and to start the planning of the Stage01 work. That planning procedure will be concluded during a later meeting in February.

Participants

Ugo Becciani, Ken Brodlie, Marco Comparato, Alessandro Costa, Claudio Gheller, James Gray, Martin Hill, Peppe Longo, Antonino Staiano, Bob Mann, Bob Nichol, Clive Page, Garry Smith, John Taylor.

Agenda

Session 1: Context

• Introduction to DS6 (Bob Mann)
  
• AstroGrid/AVO infrastructure (John Taylor)
  

Session 2: DS6 interests and plans

• Playing in high dimensions (Bob Nichol)
  
• Neural tools for data mining in massive datasets (Peppe Longo)
  
• Cross-matching catalogues and column-based storage for data exploring (Clive Page)
  
• VisIVO - a VO-enabled tool for scientific visualization and data analysis (Ugo Becciani, Claudio Gheller, Marco Comparato)
  
• Visualization for VOTech: Visualization@Leeds, multivariate visualization (Ken Brodlie)
  
• DS6@Edinburgh (Bob Mann)
  

Session 3: General Discussion

Close

Workshop Report

This workshop succeeded in its main goal, of giving the DS6 team the chance to get to know each other and each other's interests in, and plans for, the DS6 work. For the most part the workshop comprised the presentations listed above, plus associated discussion, but there was some general discussion on the scope of DS6 and how it should work.

Agreement was reached on several general points:

• A two-stream approach. DS6 Stage01 workplans should comprise two streams: one with achievable short-term goals, which will feed into the Year 1 software release; and a slower, more research-oriented track which will yield results in the longer run.
• Software survey. As part of the preparation for the Data Exploration Study Report, to be delivered in Month 15 of the project, the DS6 team should start a DS6SoftwareSurvey wiki page, on which to record assessments of relevant, existing software packages.
• Science-driven work. The work of DS6 must be science-driven. We must seek engagement with the VOTech Science Advisory Group (Science Team) when it gets up and running, but we must also capitalise on our own strengths and interests; clearly, we cannot simply build tools to solve our own scientific problems, but we are more likely to be make progress if we are strongly motivated by personal interest. This is a topic to which we should return in the DS6 Stage01 Planning Meeting, when we can get input from members of the VOTech management.

A consensus could not be reached on some other points, to which we should return in the DS6 Stage01 Planning Meeting, namely:

• Test data sets. DS6 must provide some sort of validation of any software that it contributes to VOTech software releases, but it is not clear exactly what this means. Should DS6 ensure that software yield scientifically meaningful/useful results, or only that it correctly implements a particular algorithm? Either way, there is clearly a need for DS6 to have some sort of suite of test data sets that can be used in validation work, and we need to sort out what they provide.
• Duplicating functionality. One tension in the scope of DS6 is between the development of specific prototypes and the identification of general requirements. For DS6 to get a reasonable picture of the future requirements on the Euro-VO infrastructure from data exploration, it is necessary that astronomers have prototype tools of powerful enough functionality that they can perform sufficiently realistic scientific analyses as to enable them to extrapolate to what they will want to do in, say, five years time. The development of sufficiently powerful tools takes a lot of effort, so, given the finite resources of DS6, one approach would be to devote all our efforts to, say, one data mining tool and one multivariate visualization tool. However, the problem with that approach would be that we might only identify the requirements on the Euro-VO infrastructure from these particular tools, rather than gaining an understanding of the requirements from the more general class of data exploration tools.

Actions

BobMann has to sort out various admin things, namely:

• get the official list of how much effort from whom is assigned to DS6 for Stage01;
• get everyone signed up the wiki, forum and email lists;
• prepare an agenda for the DS6 Stage01 Planning Meeting;
• draft some text defining the Scope of DS6, to be approved by the VOTech Board before the DS6 Stage01 Planning Meeting.

In preparation for the DS6 Stage01 Planning Meeting, the different members of the DS6 team have different things to think about:

• 1. The astronomers should think about a scientific scenario to use as the basis for the first DS6 deliverables. It should require both data mining and visualization, and must have some iterative/interactive step(s).
• 2. The tool makers should reflect on what they have heard about the AVO/AstroGrid infrastructure (especially MySpace, the Common Execution Architecture, and workflow/job execution) and think what would have to be done to get their existing tools working within it.
• 3. The infrastructure people should do the inverse of 2, and identify what is needed to integrate existing tools from the DS6 team described at this meeting into the VO and Grid infrastructure as it now stands.

-- BobMann - 03 Feb 2005