Scaling the “memory wall” for multi-dimensional seismic processing with algebraic compression on cerebras cs-2 systems

H Ltaief, Y Hong, L Wilson, M Jacquelin… - Proceedings of the …, 2023 - dl.acm.org
We exploit the high memory bandwidth of AI-customized Cerebras CS-2 systems for seismic
processing. By leveraging low-rank matrix approximation, we fit memory-hungry seismic …

Stochastic multi-dimensional deconvolution

M Ravasi, T Selvan, N Luiken - IEEE Transactions on …, 2022 - ieeexplore.ieee.org
Geophysical measurements such as seismic datasets contain valuable information that
originate from areas of interest in the subsurface; these seismic reflections are, however …

Target-oriented high-resolution elastic full-waveform inversion with an elastic redatuming method

Y Li, T Alkhalifah - Geophysics, 2022 - library.seg.org
Characterizing the elastic properties in deep-buried reservoirs beneath complex overburden
structures remains challenging for seismic inversion. Elastic full-waveform inversion (FWI) is …

Frequency-domain convolutive bounded component analysis algorithm for the blind separation of dependent sources

X Luo, Z Zhang, T Gong - IEEE Transactions on Instrumentation …, 2023 - ieeexplore.ieee.org
Aiming at the problem of dependent source separation in complex mechanical systems, the
highly universal frequency-domain convolutive bounded component analysis (FDCBCA) …

A physics-aware, low-rank regularization for multidimensional deconvolution

F Chen, M Ravasi, D Keyes - arXiv preprint arXiv:2312.11004, 2023 - arxiv.org
This paper presents a novel factorization-based, low-rank regularization method for solving
multidimensional deconvolution problems in the frequency domain. In this approach, each …

Steering customized ai architectures for hpc scientific applications

H Ltaief, Y Hong, A Dabah, R Alomairy… - … Conference on High …, 2023 - Springer
AI hardware technologies have revolutionized computational science. While they have been
mostly used to accelerate deep learning training and inference models for machine learning …

Solving multi-dimensional deconvolution via a nuclear-norm regularized least-squares approach

F Chen, M Ravasi, D Keyes - 84th EAGE Annual Conference & …, 2023 - earthdoc.org
Summary Multi-dimensional deconvolution (MDD), a data processing technique stemming
from the Green's function representation theorem, is commonly solved as a linear least …

Data-driven suppression of short-period multiples from laterally varying thin-layered overburden structures

H Peng, M Dukalski, P Elison, I Vasconcelos - Geophysics, 2023 - library.seg.org
Marchenko multiple elimination methods remove all orders of overburden-generated
internal multiples in a data-driven way. In the presence of thin beds, however, these …

Large-scale Marchenko imaging with distance-aware matrix reordering, tile low-rank compression, and mixed-precision computations

M Ravasi, Y Hong, H Ltaief, D Keyes… - … International Meeting for …, 2022 - library.seg.org
A variety of wave-equation-based seismic processing algorithms rely on the repeated
application of the Multi-Dimensional Convolution (MDC) operator. For large-scale 3D …

MicroDeblur: Image Motion Deblurring on Microcontroller-based Vision Systems

S Lee - Proceedings of the 22nd International Conference on …, 2023 - dl.acm.org
This paper introduces MicroDeblur, an on-device image motion deblur solution for resource-
constrained microcontroller-based vision systems. Although motion blurs caused by the …