Related papers: ELT-scale Adaptive Optics real-time control with t…
Large language models (LLMs) have been widely applied but face challenges in efficient inference. While quantization methods reduce computational demands, ultra-low bit quantization with arbitrary precision is hindered by limited GPU Tensor…
Mini-batch inference of Graph Neural Networks (GNNs) is a key problem in many real-world applications. Recently, a GNN design principle of model depth-receptive field decoupling has been proposed to address the well-known issue of…
Ternary quantization has emerged as a powerful technique for reducing both computational and memory footprint of large language models (LLM), enabling efficient real-time inference deployment without significantly compromising model…
This work describes the challenges presented by porting parts ofthe Gysela code to the Intel Xeon Phi coprocessor, as well as techniques used for optimization, vectorization and tuning that can be applied to other applications. We evaluate…
The Santa Cruz Extreme AO Lab (SEAL) is a new visible-wavelength testbed designed to advance the state of the art in wavefront control for high contrast imaging on large, segmented, ground-based telescopes. SEAL provides multiple options…
Sparse matrix vector multiplication (SpMV) is one of the most common operations in scientific and high-performance applications, and is often responsible for the application performance bottleneck. While the sparse matrix representation has…
Next-generation solar spectrographs increasingly record dense wavelength windows in which tens to hundreds of spectral lines are sampled at each spatial location and time step. This expands the scope for multi-line, multi-height diagnostics…
Adaptive optics (AO) have been used to correct wavefronts to achieve diffraction limited point spread functions in a broad range of optical applications, prominently ground-based astronomical telescopes operating in near infra-red. While…
Kernel-phase is a recently developed paradigm that tackles the classical problem of image deconvolution, based on an interferometric point of view of image formation. Kernel-phase inherits and borrows from the notion of closure-phase,…
While vision transformers (ViTs) have continuously achieved new milestones in the field of computer vision, their sophisticated network architectures with high computation and memory costs have impeded their deployment on resource-limited…
Low earth orbit (LEO) satellite-assisted communications have been considered as one of key elements in beyond 5G systems to provide wide coverage and cost-efficient data services. Such dynamic space-terrestrial topologies impose exponential…
Adaptive optics systems are an essential technology for the modern astronomy for ground based telescopes. One of the most recent revolution in the field is the introduction of the pyramid wavefront sensor. The higher performance of this…
Direct imaging and spectral characterization of exoplanets using extreme adaptive optics (ExAO) is a key science goal of future extremely large telescopes and space observatories. However, quasi-static wavefront errors will limit the…
Optical satellite-to-ground communication (OSGC) has the potential to improve access to fast and affordable Internet in remote regions. Atmospheric turbulence, however, distorts the optical beam, eroding the data rate potential when…
Modern deep learning enabled artificial neural networks, such as Deep Neural Network (DNN) and Convolutional Neural Network (CNN), have achieved a series of breaking records on a broad spectrum of recognition applications. However, the…
Spiking Neural Networks (SNNs) offer a promising solution to the problem of increasing computational and energy requirements for modern Machine Learning (ML) applications. Due to their unique data representation choice of using spikes and…
Future generations of radio interferometers targeting the 21\,cm signal at cosmological distances with $N\gg 1000$ antennas could face a significant computational challenge in building correlators with the traditional architecture, whose…
Machine learning (ML) tools such as encoder-decoder convolutional neural networks (CNN) can represent incredibly complex nonlinear functions which map between combinations of images and scalars. For example, CNNs can be used to map…
Recurrent Neural Networks (RNNs) are vital for sequential data processing. Long Short-Term Memory Autoencoders (LSTM-AEs) are particularly effective for unsupervised anomaly detection in time-series data. However, inherent sequential…
Adaptive optics systems are essential on all large telescopes where image quality is important. These are complex systems with many design parameters requiring optimisation before good performance can be achieved. The simulation of adaptive…