In low-earth orbit (LEO) satellite communications, it is difficult to obtain accurate channel estimation results in the classical time-frequency (TF) domain due to the large Doppler shift. Orthogonal time frequency space (OTFS) modulation is a potential solution since it can convert the time-variant channels in the TF domain into the time-invariant channels in the delay-Doppler (DD) domain. In this paper, we focus on multi-user random access in the LEO satellite communication system, where multiple users with identical pilot sequences will significantly reduce the user detection rate and CE accuracy. Therefore, we propose an OTFS pilot design based on the delay and Doppler spread to ensure the orthogonality of the received pilots. Additionally, we propose a channel estimation (CE) algorithm based on the sparsity of the channel in the delay-Doppler-angular domain to improve multi-user separation accuracy. Simulation results illustrate that the proposed algorithm can effectively distinguish multiple users with identical pilot sequences and improve channel estimation accuracy.

Stencil computations are widely used in high performance computing (HPC) applications. Many HPC platforms utilize the high computation capability of GPUs to accelerate stencil computations. In recent years, stencils have become more diverse in terms of stencil order, memory accesses and computation patterns. To adapt diverse stencils to GPUs, a variety of optimization techniques have been proposed. Due to the diversity of stencil patterns and GPU architectures, no single optimization technique fits all stencils. Therefore, stencil auto-tuning mechanisms have been proposed to conduct parameter search for a given combination of optimization techniques. However, parameter search for an inappropriate optimization combination (OC) misses the globally optimal solution. To address the above problems, we proposeGSTuner, an adaptive auto-tuning framework that efficiently determines the optimal parameter setting of the global optimization space for stencils on GPUs. Specifically,GSTunerrepresents stencil patterns as neighboring features and unifies feature vectors of OCs through data pre-processing. In addition,GSTunersamples parameter settings from superior OCs via the quota-based reward policy and regression mechanisms. After that,GSTuneremploys the genetic algorithm that considers sub-population similarity to reduce the cost of evolutionary search. The experiment results show thatGSTunercan identify better performing settings with higher auto-tuning speed compared to the state-of-the-art works.

Stencil computations are widely used in high performance computing (HPC) applications. Many HPC platforms utilize the high computation capability of GPUs to accelerate stencil computations. In recent years, stencils have become more diverse in terms of stencil order, memory accesses and computation patterns. To adapt diverse stencils to GPUs, a variety of optimization techniques have been proposed such as streaming and retiming. However, due to the diversity of stencil patterns and GPU architectures, no single optimization technique fits all stencils. Besides, it is challenging to choose the most cost-efficient GPU for accelerating target stencils. To address the above problems, we propose StencilMART, an automatic optimization selection framework that predicts the best optimization combination and execution time under a certain parameter setting for stencils on GPUs. Specifically, the StencilMART represents the stencil patterns as binary tensors and neighboring features through tensor assignment and feature extraction. In addition, the StencilMART implements various machine learning methods such as classification and regression that utilize stencil representation and hardware characteristics for execution time prediction. The experiment results show that the StencilMART can achieve accurate optimization selection and performance prediction for various stencils across GPUs.

The computational patterns of stencil operations are commonly used in HPC applications. Many HPC platforms utilize the computation capability of GPUs to accelerate stencil operations. In recent years, stencils have become more complex in terms of stencil order, memory accesses, and operator patterns. To adapt complex stencils to GPUs, various optimization techniques have been proposed such as blocking and unrolling. However, due to the complexity of GPU architecture, no single parameter setting of the optimization techniques fits all stencils. To address this problem, we propose csTuner, a scalable auto-tuning framework that quickly determines the optimal parameter setting for a given combination of optimization techniques. Specifically, csTuner leverages a set of statistics and machine learning methods to generate parameter groups and sampled parameter settings from the search space. In addition, csTuner adopts the genetic algorithm with approximation to reduce the cost of evolutionary search. The experimental results show that csTuner can find better performing settings with higher auto-tuning speed compared to the state-of-the-art works.

As an effective method, feature selection can reduce computational complexity and improve classification performance. A number of criteria exist for feature selection using labeled data, unlabeled data and pairwise constraints, most of which are based on the Euclidean distance. In this paper, we propose a filter method for feature selection with pairwise constraints, aiming to jointly evaluate a feature subset based on metric learning. Two criteria are designed based on the well-known Kullback-Leibler divergence for measuring the difference between must-link constraints and cannot-link constraints that can indicate the feature subset discrimination based on Keep It Simple and Straightforward (KISS) metric learning and Cross-view Quadratic Discriminant Analysis (XQDA) metric learning. To address the challenging feature selection problem, we formulate a sequential search algorithm guided by indicators that are simplified from the proposed criteria. Furthermore, we conducted several experiments on sleep staging based on electroencephalogram (EEG) recordings from the Sleep-EDF Database Expanded. The experimental results demonstrate the effectiveness of the proposed method compared with nine representative feature selection methods. On the data set from healthy volunteers and the data set from volunteers that had mild difficulty falling asleep, the classification average accuracies achieve 97.66% and 93.57% by using the proposed method, respectively.