Fiscal Year:
2011
Title:
A Rapid Model Fitting Tool Suite
Agency:
NASA
Contract:
NNX11CH10P
Award Amount:
$99,956.00
Abstract:
Instruments flown on board NASA missions often do not measure quantitiesof interest to scientists directly, but rather observable quantities. In addition,instruments often introduce artifacts into the data and techniques are needed toremove them. In both cases, scientifically relevant information can beinferred by simulating a parameterized model and adjusting the parameters tobest match observed data. This minimization process of "fitting" a model todata is computationally demanding, since it requires evaluating the often complexmodels many times and minimizing a function of many variables. Fast, more efficientfitting techniques are therefore required in order to fully harvest the scientificoutput from NASA missions.In this project, we will develop a fast, general-purpose parameter fitting software
tool suite taking advantage of inexpensive, high-performance graphics processing
units (GPUs). The goal of Phase I of this project is to demonstrate the feasibility
of model fitting on GPUs. We will therefore prototype forward models, relevant
mathematical operators and the actual parameter adjustment algorithm on GPUs
and compare their performance to a pure CPU implementation. We will particularly
focus on ease of use for the scientist to migrate their applications to this infrastructure.
During Phase I, the TRL will increase from 3 to 4.
In the Phase II, we will then harden the prototypes developed in Phase I and provide alternative modeling algorithms to support a broader range of applications. In addition, we will work with scientists from various NASA missions, including the Stratospheric Aerosol and Gas Experiment (SAGE) III mission and the Solar Dynamics Observatory (SDO) to accelerate their modeling needs. In addition to NASA missions, other federal or commercial applications
requiring to adjust model parameters to fit large sets of observed data will benefit from this project.
Business Contact:
Leslie W. Rosczyk
Business Official
7209741855
lwr@txcorp.com
Small Business Information at Submission:
Tech-X Corporation
5621 Arapahoe Ave, Suite A Boulder, CO -
EIN/Tax ID:
841256533
DUNS:
N/A
Number of Employees:
Woman-Owned:
No
Minority-Owned:
No
HUBZone-Owned:
No