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 applications onto memory-austere SRAM wafer-scale hardware, thus addressing a challenge arising in many wave-equation-based algorithms that rely on Multi-Dimensional Convolution (MDC) operators. Exploiting sparsity inherent in seismic data in the frequency domain, we implement embarrassingly parallel tile low-rank matrix-vector multiplications (TLR-MVM), which account for most of the elapsed time in MDC operations, to successfully solve the Multi-Dimensional Deconvolution (MDD) inverse problem. By reducing memory footprint along with arithmetic complexity, we fit a standard seismic benchmark dataset into the small local memories of Cerebras processing elements. Deploying TLR-MVM execution onto 48 CS-2 systems in support of MDD gives a sustained memory bandwidth of 92.58PB/s on 35, 784, 000 processing elements, a significant milestone that highlights the capabilities of AI-customized architectures to enable a new generation of seismic algorithms that will empower multiple technologies of our low-carbon future.

Seismic deblending is an ill-posed inverse problem that involves counteracting the effect of a blending matrix derived from the shots position and firing time. In this letter, we propose a seismic deblending method based on so-called deep preconditioners. A convolutional Autoencoder (AE) is first trained in a patch-wise fashion to learn an effective sparse representation of the common receiver gathers (CRGs) we aim to reconstruct. Then, the decoder branch of the trained AE is used as a nonlinear preconditioner for the deblending problem. Particularly, to avoid the explicit creation of a training dataset, we suggest to use the common shot gathers (CSGs) of the blended dataset itself to train the AE network, as they are not affected by incoherent blending noise. Numerical examples on synthetic and field datasets demonstrate the effectiveness of the proposed method in comparison with significantly comparable techniques: a dictionary-learning based deblending method; an end-to-end deblending convolution neutral network (CNN).

Geophysical measurements such as seismic datasets contain valuable information that originate from areas of interest in the subsurface; these seismic reflections are however inevitably contaminated by other events created by waves reverberating in the overburden. Multi-Dimensional Deconvolution (MDD) is a powerful technique used at various stages of the seismic processing sequence to create ideal datasets deprived of such overburden effects. Whilst the underlying forward problem holds for a single source, a successful inversion of the MDD equations requires availability of a large number of sources alongside prior information, possibly introduced in the form of physical constraints (e.g., reciprocity and causality). In this work, we present a novel formulation of time-domain MDD based on a finite-sum functional. The associated inverse problem is then solved by means of stochastic gradient descent algorithms, where the gradients at each iteration are computed using a small subset of randomly selected sources. Through synthetic and field data examples, we show that the proposed method converges more stably than the conventional approach based on full gradients. Stochastic MDD represents a novel, efficient, and robust strategy to deconvolve seismic wavefields in a multi-dimensional fashion.