With the widespread photovoltaic deployment to achieve net-zero energy goal, resulting waste draws attention. In China, considerable steps have not been taken for management. The lack of relevant scientific information on brings difficulties establishment regulatory systems. this study, necessity and feasibility recovery were investigated. stream was quantified as 48.67–60.78 million t in 2050. waste, indium, selenium, cadmium, gallium high risk, judging by metal criticality analysis, which...

The differential evolution is a floating-point encoded evolutionary algorithm for global optimization over continuous spaces. This so far uses empirically chosen fixed search parameters. study to make the more responsive changes in problem. paper proposes new adaptive form of DE having lower number parameters required be set by user priori. fuzzy logic controllers whose inputs incorporate relative function values and individuals successive generations adapt mutation operation crossover...

Sparse matrix vector multiplication (SpMV) is an important computational kernel in traditional high-performance computing and emerging data-intensive applications. Previous SpMV libraries are optimized by either application-specific or architecture-specific approaches but present difficulties for use real In this work, we develop auto-tuning system (SMATER) to bridge the gap between specific optimizations general-purpose use. SMATER provides programmers a unified interface based on...

Purpose The purpose of this paper is to improve the existing differential evolution (DE) mutation operator so as accelerate its convergence. Design/methodology/approach A new general donor form for operation in DE presented, which defines a convex combination triplet individuals selected mutation. Three schemes from that are deduced. Findings three were empirically compared with original version and variants by using suite nine well‐known test functions, also demonstrated practical...

In determining the orbits of low Earth orbit (LEO) satellites using spaceborne GPS, errors caused by receiver antenna phase center offset (PCO) and variations (PCVs) are gradually becoming a major limiting factor for continued improvements to accuracy. Shiyan 3, small satellite mission space technology experimentation climate exploration, was developed China launched on November 5, 2008. The dual-frequency GPS payload delivers 1 Hz data provides basis precise determination within range few...

General sparse matrix-matrix multiplication (SpGEMM) is an essential building block in a number of applications. In our work, we fully utilize GPU registers and shared memory to implement efficient load balanced SpGEMM comparison with the existing implementations.

Main observation and conclusion An HPLC‐UV‐guided separation was performed four pairs of unprecedented macathiohydantoin dimers, lepithiohydimerins A—D ( 1 — 4 ) bearing a rare disulfide bond were isolated from the tubers Maca. Their structures unambiguously confirmed by NMR spectroscopic, X‐ray crystallographic electronic circular dichroism (ECD) analyses. At concentration 20 μmol/L, compounds 2‐1 , 2‐2 4‐1 increased viability PC12 cells with cell at (72.06 ± 1.14)%, (72.64 1.49)%, (70.93...

The development of science and technology, especially the rise augmented reality (AR) technology provides a new way for inheritance innovation traditional intangible cultural heritage. Based on design research tide play blind box AR digital empowering heritage, this paper discusses application feasibility in analyzes strategy implementing puts forward corresponding countermeasures challenges that may be faced, aiming at providing ideas methods heritage modern technology.

The Differential Evolution (DE) is a floating-point encoded evolutionary strategy for global optimization. It has been demonstrated to be an efficient, effective, and robust optimization method, especially problems containing continuous variables. This paper concerns applying DE-based algorithm training Radial Basis Function (RBF) networks with variables including centres, weights, widths of RBFs. proposed consists three steps: the first step initial tuning, which focuses on searching...

Graph embedding training models access parameters sparsely in a “one-hot” manner. Currently, the distributed graph neural network is learned by data parallel with parameter server, which suffers significant performance and scalability problems. In this article, we analyze problems characteristics of kind on GPU clusters for first time, find that fixed model scattered among different machine nodes are major limiting factor efficiency. Based our observation, develop an efficient system called...

Variational autoencoder is one of the deep latent space generation models, which has become increasingly popular in image and anomaly detection recent years. In this paper, we first review development research status traditional variational autoencoders their variants, summarize compare performance all autoencoders. then give a possible direction VAE.

The complete mitochondrial genome of Hydra vulgaris (Hydroida: Hydridae) is composed two linear DNA molecules. (mtDNA) molecule 1 8010 bp long and contains six protein-coding genes, large subunit rRNA, methionine tryptophan tRNAs, pseudogenes consisting respectively a partial copy COI, terminal sequences at ends the mtDNA, while mtDNA 2 7576 seven small tRNA, pseudogene COI mtDNA. gene begins with GTG as start codon, whereas other 12 genes typical ATG initiation codon. In addition, all are...

Convolution computing is one of the primary time consuming part convolutional neural networks (CNNs). State art use samll, 3 × filters. Recent work on Winograd convolution can reduce computational complexity a lot, making fast. But existing implementations limited to small tiles, i.e. F(4 4, 3) and F(2 2, where 4 2 are tile sizes output channels filter size, single precision data. In this paper, we propose an optimized mixed F(6 6, implementation NVIDIA Ampere GPUs using Tensor Cores. Our...

Graph neural networks (GNNs) are prevalent to deal with graph-structured datasets, encoding graph data into low dimensional vectors. In this paper, we present a fast training network framework, i.e., WholeGraph, based on multi-GPU distributed shared memory architecture. Whole-Graph consists of partitioning the and corresponding node or edge features multi-GPUs, eliminating bottleneck communication between CPU GPUs during process. And different is implemented by GPUDirect Peer-to-Peer (P2P)...

Although explainable artificial intelligence (XAI) has achieved remarkable developments in recent years, there are few efforts have been devoted to the following problems, namely, i) how develop an method that could explain black-box a model-agnostic way? and ii) improve performance interpretability of using such explanations instead pre-collected important attributions? To explore potential solution, we propose explanation termed as Sparse Contrastive Coding (SCC) verify its effectiveness...

Five new thiohydantoin derivatives (1–5) were isolated from the rhizomes of Lepidium meyenii Walp. NMR (1H and 13C NMR, 1H−1H COSY, HSQC, HMBC), HRESIMS, ECD employed for structure elucidation compounds. Significantly, compound 1 was first example thiohydantoins with thioxohexahydroimidazo [1,5-a] pyridine moiety. Additionally, compounds 2 3 possess rare disulfide bonds. Except 4, all isolates assessed neuroprotective activities in corticosterone (CORT)-stimulated PC12 cell damage. Among...

This paper concerns the application of differential evolution to training radial basis function networks. The algorithm consists initial tuning, local and global tuning. last two tunings both use a cycle-increased searching scheme, tuning employs fuzzy adaptive control. mean square error from desired actual outputs is applied as objective function. Four standard test functions used for demonstration. A comparison net performances with approaches reported in literature shows resulting network...

Abstract Motivation Identifying the drug-protein interactions (DPIs) is crucial in drug discovery, and a number of machine learning methods have been developed to predict DPIs. Existing usually use unrealistic datasets with hidden bias, which will limit accuracy virtual screening methods. Meanwhile, most DPIs prediction pay more attention molecular representation but lack effective research on protein high-level associations between different instances. To this end, we presented here novel...