A5001868468- Scientific Computing and Data Management
- Seismic Imaging and Inversion Techniques
- Distributed and Parallel Computing Systems
- Hydraulic Fracturing and Reservoir Analysis
- Cloud Computing and Resource Management
- Advanced Data Storage Technologies
- NMR spectroscopy and applications
- Underwater Acoustics Research
- Seismic Waves and Analysis
- Drilling and Well Engineering
University of Massachusetts Amherst
2023
The University of Texas at Dallas
2019-2022
Bayesian methods for full-waveform inversion allow quantification of uncertainty in the solution, including determination interval estimates and posterior distributions model unknowns. Markov chain Monte Carlo (MCMC) produce subject to fewer assumptions, such as normality, than deterministic methods. However, MCMC is computationally a very expensive process that requires repeated solution wave equation different velocity samples. Ultimately, large proportion these samples (often 40%–90%)...
Bayesian seismic inversion can be used to sample from the posterior distribution of velocity field, thus allowing for uncertainty quantification. However, techniques like Markov chain Monte Carlo (McMC) extremely computationally expensive. We propose a two-stage McMC method where an upscaled wave equation solver is quickly filter out unacceptable fields, reducing computational expense. have found that data generated by full fine-grid correlated with solver, so valid filter. Presentation...
Monte Carlo techniques for seismic inversion are effective at producing high-quality uncertainty information about model parameters while reducing the number of assumptions required by non-stochastic full waveform algorithms. However, computationally expensive since they require repeated solution wave equation each proposed velocity model. One approach to reduce computational expense is two-stage Markov chain (MCMC), wherein a inexpensive filter used quickly reject unacceptable models. This...
Continuous integration and deployment (CICD) are fundamental to modern software development. While many platforms such as GitHub Atlassian provide cloud solutions for CICD, these don't fully meet the unique needs of high performance computing (HPC) applications. These include, but not limited to, testing distributed memory scaling studies, both which require an HPC environment. We propose a novel framework running CICD workflows on supercomputing resources. Our directly integrates with...
Linux container technologies such as Docker and Singularity offer encapsulated environments for easy execution of software. In high performance computing, this is especially important evolving complex software stacks with conflicting dependencies that must co-exist. Registry HPC (“shpc”) was created an effort to install containers in environment modules, seamlessly allowing typically hidden executables inside be presented the user commands, significantly simplifying experience. A...
Linux container technologies such as Docker and Singularity offer encapsulated environments for easy execution of software. In high performance computing, this is especially important evolving complex software stacks with conflicting dependencies that must co-exist. Registry HPC ("shpc") was created an effort to install containers in environment modules, seamlessly allowing typically hidden executables inside be presented the user commands, significantly simplifying experience. A remaining...