Gary Horne

Data farming: discovering surprise

The development of models and analysis of modeling results usually requires that models be run many times. Very few modelers are satisfied with the computing resources available to do sensitivity studies, validation and verification, measurement of effectiveness analysis, and related necessary activities. Fortunately, high performance computing, in the form of distributed computing capabilities and commodity node systems, is becoming more pervasive and cost effective. In this paper the authors describe the concept and methods of data farming, the study and development of methods, interfaces, and tools that make high performance computing readily available to modelers and allows analysts to explore the vast amount of data that results from exercising models.

Data farming around the world overview

Data farming combines the rapid prototyping capability inherent in certain simulation models with the exploratory power of high performance computing to rapidly generate insight into questions. The data farming process focuses on a more complete landscape of possible system responses, rather than attempting to pinpoint an answer. Data farming allows decision makers to more fully understand the landscape of possibilities and also consider outliers that may be discovered. Over the past decade, an international community has formed around these ideas. In 2008, International Data Farming Workshop 16 took place in Monterey, California, USA and workshop number 17 was held in Garmisch Partenkirchen, Germany. In addition to a summary of these two workshops, this paper will present an overview of the process that has developed to include the development of both methods and applications in the international data farming community.

Using agent models and data farming to explore network centric operations

Network centric operations are difficult to quantify in many respects with models or other methods. Data farming is a methodology and capability that makes use of high performance computing to run models many times. In the case of agent-based models that are relatively small, many runs can be performed in a short period of time. This capability gives modelers and their clients the enhanced ability to discover trends and outliers in results in a variety of areas. In this paper the authors discuss some notional efforts to begin to explore questions in the area of network centric operations using the agent model MANA and data farming. By observing the network behaviors and the output for traditional and information age warfare, we have created comparisons that illustrate when networked forces outfight non-networked forces

Data farming in support of nato operations: methodology and proof-of-concept

Data Farming is a process that has been developed to support decision-makers by answering questions that are not currently addressed. It uses an inter-disciplinary approach that includes modeling and simulation, high performance computing and statistical analysis to examine questions of interest with large number of alternatives. Data Farming allows for the examination of uncertain events with numerous possible outcomes and provides the capability of executing enough experiments so that both overall and unexpected results may be captured and examined for insights. In 2010, the NATO Science and Technology Organization started the three-year Task Group “Data Farming in Support of NATO” to assess and document the data farming methodology to be used for decision support. Two case studies were performed as proof-of-concept explorations to demonstrate the power of Data Farming. The paper describes the Data Farming methodology as an iterative process and summarizes the results of the case studies.

Simulation environment to assess technology insertion impact and optimized manning

The reduction in life-cycle costs for naval vessels is critical for operating a cost efficient and robust navy. Computer based simulations are an effective tool for human system integration optimization, as well as for studying the risks associated with complex interaction between crew and systems. The proposed modular simulation environment empowers analysts to choose and integrate the best combination of agent, discrete event, and physics based simulations to address questions of manning. The environment embraces advances in complexity theory for simulating non-linear systems, knowledge discovery for data analysis and distributed computing for execution environment.

Data Farming Process and Initial Network Analysis Capabilities

Data Farming, network applications and approaches to integrate network analysis and processes to the data farming paradigm are presented as approaches to address complex system questions. Data Farming is a quantified approach that examines questions in large possibility spaces using modeling and simulation. It evaluates whole landscapes of outcomes to draw insights from outcome distributions and outliers. Social network analysis and graph theory are widely used techniques for the evaluation of social systems. Incorporation of these techniques into the data farming process provides analysts examining complex systems with a powerful new suite of tools for more fully exploring and understanding the effect of interactions in complex systems. The integration of network analysis with data farming techniques provides modelers with the capability to gain insight into the effect of network attributes, whether the network is explicitly defined or emergent, on the breadth of the model outcome space and the effect of model inputs on the resultant network statistics.