Sign up & Download
Sign in

WINGS: Intelligent Workflow-Based Design of Computational Experiments

by Yolanda Gil, Varun Ratnakar, Jihie Kim, Pedro González-Calero, Paul Groth, Joshua Moody, Ewa Deelman
IEEE Intelligent Systems (2010)

Abstract

Scientists use computational experiments to study natural phenomena through the lens of software tools and computer programs. A computational experiment consists of a description of how selected datasets are to be processed by a series of software components, in what order, and with what parameter configurations. Designing computational experiments for scientific analysis is a challenging task for scientists. We present a novel approach to explore the space of possible experiments efficiently starting from varying degrees of detail and varying amounts of constraints. Our approach represents computational experiments as workflows and exploits semantic representations of datasets and software components to reason about the experiment design space. We show how the Wings workflow system can automatically track constraints, ultimately enabling scientists to focus on the core aspects of their experiments and goals.

Cite this document (BETA)

Available from ieeexplore.ieee.org
Page 1
hidden

WINGS: Intelligent Workflow-Based Design of Computational Experiments

WINGS: Intelligent Workflow­Based Design  of Computational Experiments  Yolanda Gil, Varun Ratnakar, Jihie Kim, Pedro Antonio González-Calero, Paul Groth, Joshua Moody, Ewa Deelman Information Sciences Institute University of Southern California December 1, 2009   Abstract Scientists  use  computational  experiments  to  study  natural phenomena through the lens of software tools and computer programs.  A computational  experiment  consists  of  a  description  of  how  selected datasets are to be processed by a series of software components, in what order, and with what parameter configurations. Designing computational experiments  is  a  challenging  task  for  scientists  because  of  the  many interacting constraints to configure the software and to process datasets appropriately.      We  present  a  novel  approach  to  explore  the  space  of possible  experiments  efficiently  starting  from  varying  degrees  of  detail and varying amounts of constraints provided by the user. Our approach represents  computational  experiments  as  workflows  and  exploits semantic representations of datasets and software components to reason about  the experiment design  space.   We show how the Wings workflow system can  automatically  track  constraints and  rule out  invalid designs, enabling scientists to focus on the core aspects of their experiments and goals.   Key words: 4.1.g. Workflow management  Computational Experiment Design Scientists use computational experiments  to study natural phenomena through the  lens  of  software  tools  and  computer  programs.    These  software  tools  can  be configured  with  diverse  settings  and  parameters,  enabling  scientists  to  explore different aspects of the phenomenon at hand.  A computational experiment consists of a description of how selected datasets are to be processed by a series of software components,  in  what  order,  and  with  what  parameter  configurations.  Earth scientists  use  computational  experiments  to  estimate  seismic  hazard  through simulations of  earthquake  forecasts. Biologists use  computational  experiments  for analysis of gene expression microarray data or molecular interaction networks and pathways.      Social  scientists  analyze  large  social  networks  to  discover  structural regularities based on mining relations among individuals.  
To appear in IEEE Intelligent Systems, 2010.

Sign up today - FREE

Mendeley saves you time finding and organizing research. Learn more

  • All your research in one place
  • Add and import papers easily
  • Access it anywhere, anytime

Start using Mendeley in seconds!

Already have an account? Sign in

Readership Statistics

11 Readers on Mendeley
by Discipline
 
 
by Academic Status
 
36% Ph.D. Student
 
18% Researcher (at an Academic Institution)
 
9% Lecturer
by Country
 
27% United States
 
27% Germany
 
9% United Kingdom