Variety Preserved Instance Weighting and Prototype Selection for Probabilistic Multiple Scope Simulations
Abstract
The goal of this project is to overcome the drawbacks of the current probabilistic simulation framework and to address the issue of the modeling which uses the big data, more specifically, to 1) investigate principles of data-driven probabilistic modeling to capture variety of instance distribution in a given data set for covering multiple scopes in a seamless manner exploring a method of instance weighting, 2) extend the instance weighing method to prototype selection preserving the instance distribution for big data setting, 3) develop a probabilistic simulation principle covering the multiple scopes without losing efficiency and accuracy, and 4) demonstrate the developed probabilistic modeling and simulation methods by applying them to real world problems, in particular to simulations of a meteorological system and its associated natural disasters.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Mar 23, 2016
- Source ID
- FA23861514008
Entities
People
- Takashi Washio
Organizations
- Air Force Office of Scientific Research
- Osaka University
- United States Air Force