Workshop Schedule
August 21, 2011 Sunday |
||||
| 08:00-8:50 | Opening ceremony | |||
|
Keynote Speech by Richard Colbaugh: Monsoons, Movies, Memes, and Genes: Combining KD and M&S for Prediction |
||||
| 8:50-10:00 |
Session 1: Knowledge Discovery and Earth Sciences
|
|||
| 10:00-10:30 | Coffee break | |||
| 10:30-11:55 |
Session 2: Knowledge Discovery and Behavior-Based, Social Modeling and Simulation
|
|||
| 11:55-12:00 | Wrap-up | |||
Panel Discussions
Panel 1: Knowledge Discovery and Weather Modeling and Simulation
Moderator: John K. Williams, National Center for Atmospheric Research
Description: The quantity of data produced by weather sensors, simulations and operational numerical models continues to grow rapidly. In fact, in many areas of atmospheric science, the amount of data has overwhelmed the ability of humans to analyze, understand and use it effectively. Our inability to fully exploit these data represents a loss of potential knowledge that could contribute to faster advances in scientific understanding, allow creation of more accurate models of environmental phenomena and permit deployment of more powerful decision support tools to benefit society. In a number of areas of weather research, knowledge discovery and data mining techniques hold the promise of helping to address these deficiencies. This panel discussion will highlight some of the recent successes of data mining and simulation techniques in weather research and will discuss some of the significant remaining challenges. It is hoped that the session will help foster a productive collaboration between experts in the field of knowledge discovery and those engaged in weather modeling and simulation research.
Panel 2: Knowledge Discovery and Behavior-Based, Social Modeling and SimulationModerator: Amy Henninger, U.S. Army, Center for Army Analysis
Description: Data compilation methods, whether generated by remotely deployed sensors collecting physically observable measurements or collaborative communications, such as blogs and online survey forms, are amassing extremely large repositories of information. Compounding the overwhelming volume of historical and streaming information, is the quick turn-around required to analyze and report the "essence" of this information. For example, health care facilities must evaluate data collected by emergency rooms within the context of the individual patient’s profile, the trend (over time and social agency) within the larger patient community, the payer’s historically-generated statistics, and the provider’s subjective, but expert, knowledge of the situation; and pattern recognition must be accomplished in the few hours prior to a treatment decision. Similarly, national security policy makers and action planners have a diversity of information, all with degrees of uncertainty about accuracy (vice precision), to be explored in a short time (hours to days) to support decisions with long-term international consequences. While models and simulations of actor behavior have a credible history of performance in specialized applications, these efforts typically require time-consuming human examination. Moreover, rarely do researchers apply statistical learning and modeling techniques understanding the outputs of large-scale models to "go back" and improve the simulation/ models from such analysis. This panel discussion will highlight some of the recent successes in knowledge discovery and behavioral modeling techniques, as well as the research challenges remaining.
Refer to the list of papers for title and author(s) of the accepted papers.