Shuffle Mechanism Research Articles

LiteWiSys: A Lightweight System for WiFi-based Dual-task Action Perception

As two important contents in WiFi-based action perception, detection and recognition require localizing motion regions from the entire temporal sequences and classifying the corresponding categories. Existing approaches, though yielding reasonably acceptable performances, are suffering from two major drawbacks: heavy empirical dependency and large computational complexity. In order to solve these issues, we develop LiteWiSys in this article, a lightweight system in an end-to-end deep learning manner to simultaneously detect and recognize WiFi-based human actions. Specifically, we assign different attentions on sub-carriers, which are then compressed to reduce noise and information redundancy. Then, LiteWiSys integrates deep separable convolution and a channel shuffle mechanism into a multi-scale convolutional backbone structure. By feature channel split, two network branches are obtained and further trained with a joint loss function for dual tasks. We collect different datasets at multi-scenes and conduct experiments to evaluate the performance of LiteWiSys. In comparison to existing WiFi sensing systems, LiteWiSys achieves promising precision with lower complexity.

Faster Metallic Surface Defect Detection Using Deep Learning with燙hannel燬huffling

Deep learning has been constantly improving in recent years, and a significant number of researchers have devoted themselves to the research of defect detection algorithms. Detection and recognition of small and complex targets is still a problem that needs to be solved. The authors of this research would like to present an improved defect detection model for detecting small and complex defect targets in steel surfaces. During steel strip production, mechanical forces and environmental factors cause surface defects of the steel strip. Therefore, the detection of such defects is key to the production of high-quality products. Moreover, surface defects of the steel strip cause great economic losses to the high-tech industry. So far, few studies have explored methods of identifying the defects, and most of the currently available algorithms are not sufficiently effective. Therefore, this study presents an improved real-time metallic surface defect detection model based on You Only Look Once (YOLOv5) specially designed for small networks. For the smaller features of the target, the conventional part is replaced with a depth-wise convolution and channel shuffle mechanism. Then assigning weights to Feature Pyramid Networks (FPN) output features and fusing them, increases feature propagation and the network’s characterization ability. The experimental results reveal that the improved proposed model outperforms other comparable models in terms of accuracy and detection time. The precision of the proposed model achieved by mAP@0.5 is 77.5% on the Northeastern University, Dataset (NEU-DET) and 70.18% on the GC10-DET datasets.

Small samples data augmentation and improved MobileNet for surface defects classification of hot-rolled steel strips

Surface defects in hot-rolled steel strips are one of the common product problems for the steel industry, which harm the product appearance, affect the corrosion and wear resistance, and shorten the product service life. The natural defect samples are sparse, category imbalanced, and expensive manual annotations. Therefore, it is crucial to study the data augmentation and classification methods for small sample surface defects. To solve the above problems and improve the accuracy and real-time performance of defect classification, we propose a random offline data augmentation algorithm (Random-CutMix) and an improved MobileNet architecture (SP-MobileNet). The Random-CutMix algorithm expands the dataset by random sampling to balance the number of each defect class. The SP-MobileNet combines the inverse residual module with the channel shuffle mechanism (CSIn-Module) and pyramid split attention (PSA) module, which facilitates cross-group information flow and improves model representation capability and generalization performance with low computational cost. The accuracy, recall, F1 score, parameter, computational complexity, and frame rate of SP-MobileNet with Random-CutMix on the X-SDD dataset were 95.97%, 95.22%, 95.46%, 6.5 M, and 0.54 G, 72 FPS, respectively. The experiment results indicate that our method outperforms the state-of-the-art methods and provides an effective trade-off between accuracy and instantaneity in actual industrial production.

An Attention-Based CoT-ResNet With Channel Shuffle Mechanism for Classification of Alzheimer's Disease Levels.

Detection of early morphological changes in the brain and early diagnosis are important for Alzheimer’s disease (AD), and high-resolution magnetic resonance imaging (MRI) can be used to help diagnose and predict the disease. In this paper, we proposed two improved ResNet algorithms that introduced the Contextual Transformer (CoT) module, group convolution, and Channel Shuffle mechanism into the traditional ResNet residual blocks. The CoT module is used to replace the 3 × 3 convolution in the residual block to enhance the feature extraction capability of the residual block, while the Channel Shuffle mechanism is used to reorganize the feature maps of different groups in the input layer to improve the communication between the feature maps from different groups. Images of 503 subjects, including 116 healthy controls (HC), 187 subjects with mild cognitive impairment (MCI), and 200 subjects with AD, were selected and collated from the ADNI database, and then, the data were pre-processed and sliced. After that, 10,060 slices were obtained and the three groups of AD, MCI and HC were classified using the improved algorithms. The experiments showed that the refined ResNet-18-based algorithm improved the top-1 accuracy by 2.06%, 0.33%, 1.82%, and 1.52% over the traditional ResNet-18 algorithm for four medical image classification tasks, namely AD: MCI, AD: HC, MCI: HC, and AD: MCI: HC, respectively. The enhanced ResNet-50-based algorithm improved the top-1 accuracy by 1.02%, 2.92%, 3.30%, and 1.31%, respectively, over the traditional ResNet-50 algorithm in four medical image classification tasks, demonstrating the effectiveness of the CoT module replacement and the inclusion of the channel shuffling mechanism, as well as the competitiveness of the improved algorithms.

CASM-AMFMNet: A Network Based on Coordinate Attention Shuffle Mechanism and Asymmetric Multi-Scale Fusion Module for Classification of Grape Leaf Diseases.

Grape disease is a significant contributory factor to the decline in grape yield, typically affecting the leaves first. Efficient identification of grape leaf diseases remains a critical unmet need. To mitigate background interference in grape leaf feature extraction and improve the ability to extract small disease spots, by combining the characteristic features of grape leaf diseases, we developed a novel method for disease recognition and classification in this study. First, Gaussian filters Sobel smooth de-noising Laplace operator (GSSL) was employed to reduce image noise and enhance the texture of grape leaves. A novel network designated coordinated attention shuffle mechanism-asymmetric multi-scale fusion module net (CASM-AMFMNet) was subsequently applied for grape leaf disease identification. CoAtNet was employed as the network backbone to improve model learning and generalization capabilities, which alleviated the problem of gradient explosion to a certain extent. The CASM-AMFMNet was further utilized to capture and target grape leaf disease areas, therefore reducing background interference. Finally, Asymmetric multi-scale fusion module (AMFM) was employed to extract multi-scale features from small disease spots on grape leaves for accurate identification of small target diseases. The experimental results based on our self-made grape leaf image dataset showed that, compared to existing methods, CASM-AMFMNet achieved an accuracy of 95.95%, F1 score of 95.78%, and mAP of 90.27%. Overall, the model and methods proposed in this report could successfully identify different diseases of grape leaves and provide a feasible scheme for deep learning to correctly recognize grape diseases during agricultural production that may be used as a reference for other crops diseases.

An alternative formation mechanism of {332}BCC twinning in metastable body-centered-cubic high entropy alloy

In this work, the origin of {332}BCC twinning was investigated in a body-centered-cubic (BCC) high entropy alloy (TiZrHf)87Ta13 (at.%). Detailed transmission electron microscopy analysis of the deformed microstructure revealed that a special ‘‘{351}’’<2¯11>α" type II twinning generates in the early deformation stage, and with further straining, the ‘‘{351}’’<2¯11>α" type II twin variant accommodates into a {351}α" type I twin variant. The resultant {351}α" type I twinning exhibits good lattice correspondence with the {332}BCC twinning and further propagates into the retained BCC grain with orientation relationship of {351}α”//(332)BCC, structuring into a {332}BCC twin-like configuration. Based on the experimental results and twinning analysis through evaluating twinning shear and atomic shuffle mechanism of this specific feature, our findings provide an alternative mechanism for the formation of {332}BCC twinning in which a special martensitic type II to type I twinning transformation mechanism took effect.

Spinodal decomposition coupled with a continuous crystal ordering in a titanium alloy

Spinodal decomposition mechanism is well-known for producing nanoscale modulated microstructure which is either sponge-like or plate-like. Our recent investigations of a Ti-Nb based titanium alloy have found two kinds of decomposition-induced transitions triggered simultaneously, a microstructural transition from the spongy-like to the plate-like and a crystal structure transition from bcc to hcp via the coupled atomic shear and shuffle mechanism along the Burgers pathway. Here focuses on thermal evolution kinetics of both transitions and their quantitative relations. Small angle neutron scattering profiles reveal that the aging treatments lead to a gradual change from an oval pattern to a butterfly pattern due to the spongy-to-plate microstructural transition. To characterize the crystal transition and its induced habit plane change, we define an ordering parameter from the shear component of the bcc-hcp transition. The results reveal that both follow the well-known power-law approximation with an activation energy of ∼209 kJ/mol, which is identical with diffusion energy of Nb in binary Ti-Nb alloy. However, their time exponential factors are different, about 1/5 for the microstructure growth and 1/16 for the crystal ordering and its induced habit plane change. The former is less than the common 1/3 law of the decomposition mechanism and the nucleation and growth mechanism. Aided by these kinetic equations, the microstructural and chemical parameters can be modeled by the crystal ordering. This provides a new strategy to tune the nanoscale microstructure by this novel spinodal decomposition coupled with a continuous crystal ordering.

A 91.0-dB SFDR Single-Coarse Dual-Fine Pipelined-SAR ADC With Split-Based Background Calibration in 28-nm CMOS

This paper presents a single-coarse dual-fine architecture that improves energy-efficiency of pipelined-SAR analog-to-digital converters (ADCs). A coarse and fast sub-ADC is used to quantize the most significant bits (MSBs), which are encoded with a proposed residue transformation method to control the residue generation of the first stages in two fine channels. The residue voltages generate on the capacitive digital-to-analog converters (C-DACs) of split fine channels directly without successive approximation processes. Therefore, the conversion rate is increased and the power is reduced. A shuffle mechanism is introduced into split-ADC based digital background calibration to avoid the divergence of the conventional algorithm in the proposed architecture. A high-energy-efficiency dynamic amplifier is also introduced as the residue amplifier. A 14-bit 60-MS/s ADC is prototyped in a 28-nm CMOS process. The digital calibration engine operates under 0.9-V supply, other parts of the ADC core operate under 1.05-V supply. The ADC core consumes 4.26 mW. Measurement results show that the calibration improved the signal-to-noise and distortion ratio (SNDR) and spur-free dynamic range (SFDR) dramatically, the calibrated ADC achieves SNDR and SFDR of 66.9 dB and of 91.0 dB respectively, translating to a Schreier FoM of 165.4 dB and a Walden FoM of 39.3 fJ/conversion-step.

Stress-induced α″ martensitic phase transformation and martensitic twinning in a metastable β titanium alloy

The stress-induced martensitic (SIM) β→α″ phase transformation and SIM α″ twinning in the fully-β Ti–7Mo–3Nb–3Cr–3Al (Ti-7333) alloy was investigated as a function of strain, following a surface-to-bulk methodology assisted by electron backscatter diffraction (EBSD), focused ion beam (FIB) milling and transmission electron microscopy (TEM). The results indicate that the Ti-7333 alloy is predominantly deformed by SIM and SIM α″ twinning. At the onset of deformation, the lattice correspondent SIM α″ martensite variant that produces the maximum transformation strain along the tensile direction is activated firstly, which follows the <-110>β//<001>α″ orientation relationship. As the strain increases, the SIM α″ laths deform by twinning, predominantly through the {130}<310>α″ compound twinning and {111}α″ type Ⅰ twinning modes. Their activation can be rationalized in terms of the magnitude of the shear and the complexity of the atomic shuffle using the Bilby-Crocker deformation twinning theory. The prevalence of {130}<310>α″ compound twinning in this alloy is attributed to the comparatively small shear (0.1872) and the simple shuffle (q = 2, ΔⅠa = 0.3257) mechanism involved. During the last stages of deformation, secondary microtwinning takes place within the primary twins due to the reduced mobility of the intervariant boundaries. The evolution of such hierarchical microstructure with strain accounts for the complex microstructural features that developed with deformation, responsible for the high work hardening displayed by this alloy.

Shuffle and glide mechanisms of prismatic dislocations in a hexagonal C14 -type Laves-phase intermetallic compound

Complex structured intermetallic compounds often show brittleness at ambient temperature or under high strain rate deformation, but ductility at elevated temperature or low strain rate deformation. The change of dislocation behavior plays a decisive role in the brittle-to-ductile transition. Here, dislocations are studied in a hexagonal structured Laves-phase intermetallic compound with aberration-corrected scanning transmission electron microscopy after quasistatic compression at high temperature or high-speed impact at room temperature. Different slip mechanisms are resolved for the prismatic dislocations. Under the high-temperature compression, a shuffle mechanism assisted by diffusion mediates the dislocation motion to form planar defects strictly following the topologically close-packing rules. In contrast, under the high-speed room-temperature impact, the prismatic dislocations prefer to glide on undulated planes that comprise small facets with high atom density.

Magnet

Over the past decade, Apache Spark has become a popular compute engine for large scale data processing. Similar to other compute engines based on the MapReduce compute paradigm, the shuffle operation, namely the all-to-all transfer of the intermediate data, plays an important role in Spark. At LinkedIn, with the rapid growth of the data size and scale of the Spark deployment, the shuffle operation is becoming a bottleneck of further scaling the infrastructure. This has led to overall job slowness and even failures for long running jobs. This not only impacts developer productivity for addressing such slowness and failures, but also results in high operational cost of infrastructure. In this work, we describe the main bottlenecks impacting shuffle scalability. We propose Magnet, a novel shuffle mechanism that can scale to handle petabytes of daily shuffled data and clusters with thousands of nodes. Magnet is designed to work with both on-prem and cloud-based cluster deployments. It addresses a key shuffle scalability bottleneck by merging fragmented intermediate shuffle data into large blocks. Magnet provides further improvements by co-locating merged blocks with the reduce tasks. Our benchmarks show that Magnet significantly improves shuffle performance independent of the underlying hardware. Magnet reduces the end-to-end runtime of Linkedln's production Spark jobs by nearly 30%. Furthermore, Magnet improves user productivity by removing the shuffle related tuning burden from users.

Strain direction dependency of deformation mechanisms in an HCP-Ti crystalline by molecular dynamics simulations

In this work, effects of uniaxial tensile directions on the deformation mechanisms of a hexagonal close-packed (HCP) titanium crystalline were investigated by molecular dynamics simulations. Three uniaxial tensile directions, namely the [21-1-0], [011-0] and [0001] directions, were studied. When the tensile loading was along the [21-1-0] direction, the parent HCP phase transformed firstly into the body-centered cubic (BCC) phase following the Pitsch-Schrader orientation relationship (OR), and then transformed either into the face-centered cubic (FCC) phase following the Bain path or back to the HCP phase following different variants of the Pitsch-Schrader OR. The new-forming and matrix HCP structures were in a {101-1} twinning relationship with each other. The newly formed FCC phase was in a prismatic-type (P-type) OR with the HCP matrix and in a basal-type (B-type) relationship with the new-forming HCP structure at individual contacting interface. The FCC/HCP interfaces in the P-type OR was immobile while that in the B-type OR propagated by the slip of Shockley partial dislocations. Both FCC/HCP interfaces in the P-type and B-type ORs were observed under high-resolution transmission electron microscope. With the tensile loading along the [011-0] direction, deformation mechanism of the system was dominated by the slip and dissociation of prismatic <a> dislocations. The system stretched along the [0001] direction deformed mainly through the slip and dissociation of pyramidal <c + a> dislocations, as well as through the activation of {101-2} twinning by a pure shuffle mechanism. The present investigation can provide clues in designing titanium alloys with both high strength and good plasticity.

An efficient attribute reduction algorithm using MapReduce

Classical attribute reduction algorithms based on attribute significance initiate too many jobs ( O(| C|2)) when they run in MapReduce. To improve the efficiencies of these algorithms, we proposed a novel reduction algorithm. Instead of focusing on attribute significance, the notion of a core attribute was applied to construct a new heuristic reduction algorithm, and only | C| jobs were considered to obtain a reduct. The algorithm only included two basic operations: compare and sort. The latter was optimised using the shuffle mechanism in MapReduce, which provided an efficient sorting ability for big data. In particular, we connected jobs in an iterative form to transfer the processing result of the former job to the latter job. Finally, experimental results demonstrated that the proposed attribute reduction algorithm was efficient and significantly improved upon the classical algorithms in runtime and number of jobs.

Investigation of a nano-scale, incommensurate, modulated domain in a Ti-Fe alloy

The microstructure of a Ti-Fe compositional gradient sample was investigated using transmission electron microscopy (TEM) and high angle annular dark field scanning transmission electron microscopy (HAADF-STEM). Different combinations of phases were observed to form depending on the local concentration of Fe. Interestingly, a nano-scale incommensurate modulated domain was detected in the alloy containing approximately 13.8 to 33.9 wt% Fe using selected area diffraction (SAD). The observed disordered structure with no fixed long range periodicity in the nano-scale incommensurate modulated domain in aberration-corrected HAADF-STEM imaging was found not related to the shuffle mechanism of the ω phase.

Atomistic simulations on the dynamic properties of shock and release melting in single crystal Al

Molecular dynamics simulations are performed to investigate the dynamic properties of melting in single crystal Al under [001] shock loading and release. Our simulations suggest that the shock melting can be accompanied with shuffle mechanism and especially with collision cascade. The overheated interval and the relaxation of shear stress below and above shock melting pressure are presented by the microstructure and dynamic paths. The Hugoniot results can describe the shocked solid, liquid and also solid-liquid phases, and the range of shock melting pressure is about 125–155 GPa. From different release paths, we observe the transition from surface melting to sustained release meting. The shock pressure leading to melt at release is about 72 GPa. For higher shock pressure, the melting speed will keep some constant after the early nucleation. In fact, the release melting involve both the solid-liquid interface movement and the continuous nucleation in front of it. Finally, we present the different release P-T paths (solid-solid, solid-liquid and liquid-liquid) to show the difference of melting near and far away from the free surface.

Atomistic simulation of the tension/compression response of textured nanocrystalline HCP Zr

Molecular dynamics virtual tension/compression tests were performed to study the deformation mechanisms in nanocrystalline (15–38nm grain size) 50 grains HCP Zr columnar samples with random [1−100] prismatic and [0001] basal textures. Two different embedded atom potentials were used to model atomic interactions and to describe the relationship between generalized stacking fault energies and mechanical predictions. Both potentials predict higher flow stresses in compression for both textures. Nanocrystalline Zr [0001] basal textured samples deform mainly by dislocations emission and glide along 〈11−20〉 {1−100} and by grain boundary sliding and migration; both potentials show the same deformation mechanisms for this texture. Prism [1−100] textured samples deform mainly by twinning. Both in tensile and compression tests {11−21} twins nucleate mainly at grain boundaries and sites with high stress concentration. These twins grow by a shuffle mechanism. Emission and sliding of partial dislocations, grain boundary gliding and migration were also observed for this texture with the main dislocations being of the 1/3 〈1−210〉 type. These basic deformation mechanisms were the same for both potentials tested. The observed quantitative differences between the predictions of both potentials are discussed in terms of some of their basic properties.

Interplay between interstitial displacement and displacive lattice transformations

Diffusionless displacive lattice rearrangements, which include martensitic transformations, are in real materials often accompanied by a displacive drag of interstitials. The interplay of both processes leads to a particular atomistic arrangement of the interstitials in the product phase, which is decisive for its performance. An archetype example is the martensitic transformation in Fe-C alloys. One of the puzzles for this system is that the deviation from the cubic symmetry (i.e., the tetragonality) in the martensite resulting from this interplay is lower than what thermodynamics dictates. In our ab initio approach, the relative motion of C in the transforming lattice is studied with the nudged elastic band method. We prove that an atomic shearlike shuffle mechanism of adjacent $(11\overline{2})$ Fe layers along the $\ifmmode\pm\else\textpm\fi{}{[111]}_{\mathrm{bcc}}$ directions is essential to achieve a redistribution of C atoms during the fcc $\ensuremath{\rightarrow}$ bcc transition, which fully explains the abnormal behavior. Furthermore, the good agreement with experiment validates our method to treat a diffusionless redistribution of interstitials and a displacive rearrangement of the host lattice simultaneously.

The effect of alloy composition on instabilities in the β phase of titanium alloys

A metastable nano-scale disordered precipitate with orthorhombic symmetry has been identified using high resolution scanning transmission electron microscopy. The phase, termed O′, is metastable, formed by a shuffle mechanism involving a 110<11̅0> transverse phonon wave in samples of Ti–26Nb–2Zr (at.%) quenched from the β phase. The addition of 2% Zr to Ti–26Nb appears to suppress significantly the stability of both the 112̅<111> shear and 2/3 <111> longitudinal phonon wave but promotes the 110<11̅0> transverse shuffle. This results in the nano-size O′ phase being homogeneously formed in the parent β phase matrix rather than the massive α″ phase.

The evolution of microstructure and mechanical properties of Ti–5Al–5Mo–5V–2Cr–1Fe during ageing

The phase transformations and compositional changes occurring during thermo-mechanical processing and subsequent high temperature ageing of Ti–5Al–5Mo–5V–2Cr–1Fe (wt.%) were investigated using scanning transmission electron microscopy (STEM) and atom probe tomography (APT). High resolution STEM revealed nano-sized α (<10nm) and athermal ω (∼1–3nm) formed during accelerated cooling from 800°C and slow heating to an ageing temperature of 650°C. Nuclei of α were found to form heterogeneously in the β matrix as well as at the ω phase. APT revealed pronounced Mo compositional fluctuations in the β matrix. No direct connection was established between Mo-rich or Mo-lean regions and α or ω nuclei. APT also failed to detect the ω phase, which supports theories that it forms by a shuffle mechanism, without any compositional difference from the β phase. Very small α particles, after initial ageing, showed only a minute change in composition with respect to the β matrix, indicative of a displacive–diffusional transformation. With further ageing, growth of the α lamellae was accompanied by compositional changes according to the diffusion rates of β-stabilising elements. Pile-up of the slowest diffusing solutes (Mo, V) at the α/β interface were pronounced in the initial stages of ageing. The best combination of mechanical properties (1200MPa ultimate tensile strength with 15% total elongation) was recorded after 3.6ks of ageing.

Anisotropic shock response of titanium: Reorientation and transformation mechanisms

We investigate shock-induced phase transformations in titanium (α-Ti) single crystals induced by shock loading along the [0001], [101¯0] and [121¯0] directions using molecular dynamics simulations. We find a significant dependence of the microstructure evolution on the crystallographic shock direction, providing insight into the nature of the coupling between deformation and phase transformation. For shock along the c-axis, the orientation relationships (ORs) (0001)α//(101¯0)ω and [101¯0]α//[112¯3]ω between parent and product phases are observed, which differs from that previously reported for Ti. For shock compression along the [101¯0] and [121¯0] directions, there is a reorientation of the hexagonally close-packed α phase before the α→ω martensitic transformation, and the OR is consistent with the previously proposed Silcock relationship. We associate the reorientation with a shuffle and shear mechanism and suggest that shear stress is an underlying factor for the anisotropic phase transformation sensitivity.