Summary of 2008 Meeting of Artificial Intelligence Committee

Sue Ellen Haupt, Chair

 

            The annual meeting of the Committee on Artificial Intelligence Applications to Environmental Science was held on Monday, January 21 in conjunction with the AMS Annual Meeting, beginning at 7:00 P.M. at Mulate’s restaurant in New Orleans, LA.  Unbeknownst to us a priori, Mulate’s has a live band and dancing, making discussion difficult except during the band’s break (which is precisely when our meals arrived).  So the meeting was held in two phases, Part I during that 18 min. break (Mike Richman negotiated the additional 3 min. with the band) and Part II in the lobby of the Ramada after dinner.  (But we had fun!!!)

 

Attendees to All or Part of the Meeting:

 

John Williams                            NCAR

John Trostel                              Georgia Tech

Philippe Tissot                           TAMU/Corpus Christi

Daniel Prouty                             TAMU/Corpus Christi

Haig Iskenderian                        MIT Lincoln Lab

Waylon Collins                          NOAA/NWS/CRP

Michael Richman                       Univ. Oklahoma

William Hsieh                            Univ. British Columbia

Vladimir Krasnopolsky              NOAA/NCEP/EMC

Francois Neville                        Univ. Maine

Bill Myers                                 NCAR

Caren Marzban                         Univ. Washington

Robyn Ball                                TAMU/Corpus Christi

Sue Haupt, Chair                       Penn State

 

Items of Business:

 

  1. Issues raised by Sue regarding the annual STAC meeting

 

    1. Our website was one of the ones highlighted as a “good” site at the STAC meeting.  Thanks go to Ron Holmes for the great job of reformulating our website.  Some ideas for further enhancements suggested at the STAC meeting include photos from the conference, particularly of students receiving awards; past meeting notes; and past history of the committee.  [Follow-up:  Sue took photos of the AI contest that can be forwarded to Ron for website insertion, he can post the final version of these notes, and Sue will work on following up the committee history as a project this year.
    2. Part of the STAC committees’ missions include nominating members for awards and AMS Fellow.  Although AI is not on any award committees, we can nominate Fellows and should.  [Follow-up:  Sue will take steps to nominate at least one Fellow this year.]
    3. There is a new venue for education – the Education Forum.  These forums are relatively new, aimed at the first year graduate student, and less expensive than a short course.  We should discuss whether this is an appropriate venue for AI.
    4. Venues for future conferences: 

                                                               i.      Probability and Statistics is considering a smaller meeting for summer of 2010 and is looking for other committees to partner with them.

                                                             ii.      The 2010 AMS Annual Meeting will be in Atlanta with a theme: Weather Climate Change.  There will be an emphasis on predicting extreme events that may be appropriate for AI committee participation.

 

  1. Open Discussion

 

    1. Should we change the committee name?  Artificial Intelligence is becoming “passé” and the entire committee name is quite long.  Several discussion items came to light:

                                                               i.      Other disciplines have moved to “fresher” names, such as Computational Intelligence, Soft Computing, Data Mining.  There was some discussion of the implications of each.

                                                             ii.      Is “Applications to Environmental Science” necessary or helpful?  Some commented that it likely came from a lack of confidence in the early committee members as to their place in AMS and we are now past that.  Several of the non-committee attendees commented that this extended name is what attracted them to our sessions and our committee.  Another comment is that it makes the committee easier to locate with a “google.”

The item was tabled - to be continued via email discussion.

    1. Should we hold a Short Course or Educational Forum next year and, if so, what is our target audience and scope?  After some discussion, we decided that the Short Course venue is more appropriate, but we should keep the course to limited scope, for one day only, and on a “hot topic” in AI (random forests was an example that was mentioned.  We will have a one day Short Course next year on a topic to be decided after email discussion.
    2. Should we have a conference next year and, if so, what will be our emphasis.  We decided that we are ready for annual conferences that emphasize joint sessions with our conferences.  We brainstormed conferences to approach for joint sessions and decided:

                                                               i.      Climate – Sue will talk to Jerry Potter and William Hsieh will coordinate.

                                                             ii.      Satellites – Sue will investigate.  Perhaps Dennis Boccippio would be an appropriate chair.

                                                            iii.      Hydrology – Philippe will chair.

                                                           iv.      Air Quality – Sue will talk to Mike Brown.

                                                             v.      Radar – John Williams will talk to their chair and chair from the AI side.

    1. Should we have another AI Contest?  Some discussion ensued about the appropriateness and what should be done differently.  Those who participated felt that it was a useful exercise.  The committee decided to resume discussion after the session on Tuesday.

 

Further Discussion after AI Contest on Tuesday January 22:

 

Much discussion ensued post-processing the AI contest.  Although there was some concern about sending wrong messages about one method being “better” than another, particularly when judged by a single metric, the group decided that it was a good thing, motivating, interesting, a good learning exercise, and fun.  We would like to make the contest a “tradition.”  It would be nice to motivate more “beginners” in AI.  Some ideas for improving the contest:

-         Pair inexperienced AI users with more experienced colleagues.

-         Lead off the contest with an extensive introduction to AI.

-         Finish off with a panel of “what learned.”

-         Choose and disseminate data earlier, perhaps by late Spring, to get more classes to participate.

-         Grow the contest by moving most of it to a poster contest.  Limit the speaking slots to “invited” speakers based on a combination of performance and an attempt to showcase a variety of methods.

-         Tie the metrics to how the data forecasts would be used.  For instance, if the NWS forecasters would be judged by a specific metric, concentrate on that.

-         Work to include “traditional” and “statistical” methods as well.

-         Judge on a probabilistic basis:  eg. ROC plots.  Set a number and try for performance beyond a given baseline.

-         Ranking is part of the motivation to do it.  The cash prize is a nice bonus.

-         Communicate via a listserve or Wiki or Google docs.

 

Other Business Handled Later:

 

  1. John Williams attended the Thursday meeting regarding planning next year’s conference.  He communicated the information to Sue who will serve as program chair.
  2. John Williams approached the chair of the Radar committee regarding a joint session in 2009.  Initial response is quite positive.
  3. Sue Haupt approached the chair of the Air Pollution committee regarding a joint session in 2009 and initial response is positive.
  4. Mike Richman and V. Lakshmanan will be in charge of the 2009 AI Competition.  They suggested finding an interesting application and asking another conference to team or to contribute a committee member.

 

Conference Attendance Summary:

 

SEH spot-counted in some sessions:

 

Special Session on Storm Surges:  about 35, sometimes more

Joint Session with IIPS: 70-87

AI Contest:  26 (beginning), 37, 41 (end)

Joint with Probability and Stats: 37 (early), more second session but didn’t count

Joint with ARAM:  66, 77

 

Action Items to complete via email and individual effort:

 

  1. Continue discussion on changing the committee name. (All)
  2. Decide on the emphasis of the 2009 Short Course. (All)
  3. Nominate a Fellow.  (Sue and volunteers)
  4. Document committee history.  (Sue and volunteers)
  5. Plan joint sessions for 2009 conference. (Sue and volunteers)
  6. Continue updating website. (Ron, Sue, contributors)