Sparse Matrix-matrix Multiplication (SpMM) and Sampled Dense-dense Matrix (SDDMM) are important sparse operators in scientific computing deep learning. Tensor Core Units (TCUs) enhance modern accelerators with superior power, which is promising to boost the performance of matrix a higher level. However, due irregularity unstructured data, it difficult deliver practical speedups on TCUs. To this end, we propose FlashSparse, novel approach bridge gap between workloads TCU architecture....

This paper presents a steering-wheel-angle-based driver fatigue detection method for real driving conditions.This extracts approximate entropy (ApEn) feature from recorded steering wheel angle (SWA) signal with decision-tree-like classifier to identify the level.ApEn is extracted fixed-size sliding window on real-time SWA series.To further exploit in-depth information of SWA, additional features including intervalpercentage, deviation, kurtosis and complexity value ApEn are applied designed...

High sensitivity detection of terahertz waves can be achieved with a graphene nanomesh as grating to improve the coupling efficiency incident and using nanostructure energy gap enhance excitation plasmon. Herein, fabrication process FET THz detector based on rectangular GNM (r-GNM) is designed, developed, including CVD growth wet-process transfer high quality monolayer films, preparation r-GNM by electron-beam lithography oxygen plasma etching, gate electrodes Si3N4 dielectric layer. The...

Abstract To improve the linear range of measurement, a restrictor is usually added to main flow channel capillary-tube-type thermal mass flowmeter. However, as flowmeter increases, singularity will occur in low velocity section calibration data, which lead significant decrease measurement accuracy this section. address issue, first, paper developed microelectromechanical system (MEMS) gas based on principle flow. Then, effect bypass ratio its performance was further investigated using...

Music emotion recognition (MER) is a challenging field of studies that has been addressed in multiple disciplines such as cognitive science, physiology, psychology, musicology, and arts. In this paper, music emotions are modeled set continuous variables composed valence arousal (VA) values based on the Valence-Arousal model. MER formulated regression problem where 548 dimensions features were extracted selected. A wide range methods including multivariate adaptive spline, support vector...

The success of convolutional neural networks (CNNs) benefits from the stacking layers, which improves model's receptive field for image data but also causes a decrease in inference speed. To improve speed large network models without sacrificing performance indicators too much, data-aware adaptive pruning algorithm is proposed. consists two parts, namely, channel method based on attention mechanism and policy reinforcement learning. Experimental results CIFAR-100 dataset show that proposed...

Sparse Matrix-matrix Multiplication (SpMM) and Sampled Dense-dense Matrix (SDDMM) are important sparse operators in scientific computing deep learning. Tensor Core Units (TCUs) enhance modern accelerators with superior power, which is promising to boost the performance of matrix a higher level. However, due irregularity unstructured data, it difficult deliver practical speedups on TCUs. To this end, we propose FlashSparse, novel approach bridge gap between workloads TCU architecture....

An intelligent street lamp management system based on GPRS (General Packet Radio Service) and ZigBee communication network is designed realized for the of urban lamps. The lamps can be controlled locally or remotely by proposed system. With local control, on/off automatically according to surrounding environment. remote manually. running status detected send upper computer time, which give information fault location. zero point detection circuit PWM (Pulse Width Modulation) control...

It is a common problem of manufacturing enterprise limited by business line, which optimized its resources allocation to increase productivity, but now the researches are very weak in relationship mapping and network analysis method world. Aimed demands, proposed resource line system based on relational production system, model hierarchical node degree been established. Defined data structure network, simplified computing dividing it three parts - orders assignment, accomplish feedback pure...

A design method of Ku-band high performance frequency synthesizer is introduced in this paper. Based on the traditional loop mixing phase locking method, influence additional noise digital detector, divider and other components output reduced through reasonable planning, optimization bias division ratio. The developed has low spurious similar to that direct synthesis obtained, which good engineering application. At same time, volume greatly by adopting 3D assembly technology RF heterogeneous...

<title>Abstract</title> The sparse triangular solver (SpTRSV) is mostly used for scientific and engineering applications. structured-grid of regular dependencies (STRSV) a special kind SpTRSV. Some general SpTRSVs that disregards the regularity STRSV are unsuitable STRSV. This paper proposes an efficient parallel algorithm on SW26010 (swStructTRSV). makes full use fine-grained low latency communication characteristics to reduce waiting time synchronization, maximizes access improve memory...

In the measuring technology of microscopy, in order to realize high focusing accuracy and also strong noise suppression, author approves a new method for microscopy on base principle microscope imaging focus.The uses Canny algorithm recognize edge measured object.When object is narrowest, gets best focusing.The experiment shows that compared with classical evaluation function, this characteristic unimodality, accuracy, sensibility immunity.The an autofucusing combination property, it...

The precision of temperature control is very important for the iron ore reduction performance test.Due to conventional PID controller can not be parameter self-tuning on line, this paper proposed that scheme based fuzzy adaptive intelligent by combining with technology.By simulating furnace heating and cooling process, it shown rate heat preservation process achieve good control.Using Matlab toolbox simulink simulation, result shows has high precision, small overshoot, dynamic characteristics.

It is very normal that the soft sensor model based on computational intelligence would adopt so many secondary variables, there redundant correlative information in input space, it easy to appear curse of dimensionality, and complexity large. In order obtain parsimonious sleep fidget high at precision low complexity, authors feature extraction methods such as principal component analysis (PCA) kernel (KPCA) decrease dimensionalities model, build five kinds for sensor. Simulation results show...

Models of armchair graphene nano ribbons (AGNR) and zigzag nanoribbons (ZGNR) are constructed the energy band these nanostructures calculated using semi empirical extended huckel molecular orbital (EHMO) tight binding method. Based on Huckel model, electronic structures were simulated analyzed ATK/VNL simulation platform. The results in good agreement with those model first principles density functional theory reported earlier, which verifies accuracy calculation structure carbon atoms by...

In the above article <xref ref-type="bibr" rid="ref1" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">[1]</xref> , unfortunately, we had missed grant number CQYC20200309237 of funding organization Chongqing Talent Plan. This affected only detail and no influence on discussion conclusions in article. The complete correct organizations with are as follows.

Reaches with very thick deposits alternate deeply incised reaches in the DaDu river valley. The sediments are described and a number of 14 C dates given. alluvium, Pleistocene slope‐wash deposits, marginal moraines beds fluvio‐lacustrine origin. valley development is explained as result neotectonic movement glacial aggradation. From point view water‐power development, three different types can be distinguished, each its own consequences for hydraulic engineering works.