1. CrimeStat (Jan 2007 - Sep 2007)
CrimeStat is a spatial statistics software for analysis of crime incident locations. I have been working on creating the next generation CrimeStat that will make the modules within CrimeStat reusable so that they can be consumed by custom applications. Technically, my job is to redesign the architecture of existing software project in C++ to create .NET based components.
1. Context-Inclusive Approach to Speed-up Function Evaluation for Statistical Queries: An Extended Abstract
Vijay Gandhi, James M. Kang, Shashi Shekhar, Junchang Ju, Eric D. Kolaczyk, Sucharita Gopal
First International Workshop on Spatial and Spatio-Temporal Data Mining, 2006 PaperSlides Abstract:
Many statistical queries such as maximum likelihood estimation involve finding the best candidate model given a set of candidate models and a quality estimation function.
This problem is common in important applications like land-use classification at multiple spatial resolutions from remote sensing raster data.
Such a problem is computationally challenging due to the significant computation cost to evaluate the quality estimation function for each candidate model.
A recently proposed method of multiscale, multigranular classification has high computational overhead of function evaluation for various candidate models independently before comparison.
In contrast, we propose a context-inclusive approach that controls the computational overhead based on the context, i.e. the value of the quality estimation function for the best candidate model so far.
Experimental results using land-use classification at multiple spatial resolutions from satellite imagery show that the proposed approach reduces the computational cost significantly while providing comparable classification accuracy.
2. Parallelizing Multiscale and Multigranular Spatial Data Mining Algorithm
Vijay Gandhi, Mete Celik, Shashi Shekhar
The Second Conference on Partitioned Global Address Space Programming Models, 2006 PaperSlides Abstract:
Multiscale and Multigranular (MSMG) Spatial Data
Mining (SDM) algorithms are used to find the best
granular class label from a hierarchical set of granular
class labels for spatial classification, which is important
for many application domains including the military.
However, it is computationally very expensive due to a
complex quality measure for ranking class labels. In this
paper we propose a parallel formulation of a MSMGSDM
algorithm to scale up to the problem sizes of interest
to the Army using the Partitioned Global Address Space
(PGAS) model programmed in Unified Parallel C (UPC),
which facilitates sharing of data among processors.
Experimental evaluations for land cover classification
from satellite imagery show that the proposed parallel
formulation achieves speedup of 6.65 using 8 processors.
The views and opinions expressed in this page are strictly those of the page author. The contents of this page have not been reviewed or approved by the University of Minnesota.