Explore the vast range of titles available.

Visual sensors, including 3D light detection and ranging, neuromorphic dynamic vision sensor, and conventional frame cameras, are increasingly integrated into edge-side intelligent machines. However, their data are heterogeneous, causing complexity in system development. Moreover, conventional digital hardware is constrained by von Neumann bottleneck and the physical limit of transistor scaling. The computational demands of training ever-growing models further exacerbate these challenges. We propose a hardware-software co-designed random resistive memory-based deep extreme point learning machine. Data-wise, the multi-sensory data are unified as point set and processed universally. Software-wise, most weights are exempted from training. Hardware-wise, nanoscale resistive memory enables collocation of memory and processing, and leverages the inherent programming stochasticity for generating random weights. The co-design system is validated on 3D segmentation (ShapeNet), event recognition (DVS128 Gesture), and image classification (Fashion-MNIST) tasks, achieving accuracy comparable to conventional systems while delivering 6.78 × /21.04 × /15.79 × energy efficiency improvements and 70.12%/89.46%/85.61% training cost reductions. Processing heterogeneous visual data in edge-side intelligent machines is complex and inefficient. Here, the authors propose a hardware-software co-designed system using random resistive memory, achieving significant energy efficiency and training cost reductions.

Intrinsic plasticity of neurons, such as spontaneous threshold lowering (STL) to modulate neuronal excitability, is key to spatial attention of biological neural systems. In‐memory computing with emerging memristors is expected to solve the memory bottleneck of the von Neumann architecture commonly used in conventional digital computers and is deemed a promising solution to this bioinspired computing paradigm. Nonetheless, conventional memristors are incapable of implementing the STL plasticity of neurons due to their first‐order dynamics. Here, a second‐order memristor is experimentally demonstrated using yttria‐stabilized zirconia with Ag doping (YSZ:Ag) that exhibits STL functionality. The physical origin of the second‐order dynamics, i.e., the size evolution of Ag nanoclusters, is uncovered through transmission electron microscopy (TEM), which is leveraged to model the STL neuron. STL‐based spatial attention in a spiking convolutional neural network (SCNN) is demonstrated, improving the accuracy of a multiobject detection task from 70% (20%) to 90% (80%) for the object within (outside) the area receiving attention. This second‐order memristor with intrinsic STL dynamics paves the way for future machine intelligence, enabling high‐efficiency, compact footprint, and hardware‐encoded plasticity. Spontaneous threshold lowering (STL) originates from the intrinsic neuronal plasticity observed in biological neurons, which plays an important role in a number of learning protocols like spatial attention. Here for the first time artificial STL neurons using second‐order yttria‐stabilized zirconia with Ag doping (YSZ:Ag) memristors at a small hardware overhead are realized, which mimic neural intrinsic plasticity and boost the performance of spiking neural networks.

Background Consistent patterns of gut microbiota variations, particularly in relative abundance, have been identified in the adult human gut. Enterotype, another general measure of the gut microbiota, is a valuable approach for categorizing the human gut microbiota into distinct clusters. The impact of different enterotypes on human health varies, and the changes induced by moderate altitude exposure remain unclear. This study aimed to conduct a comprehensive investigation of the cascade effects triggered by enterotype shifts following moderate altitude exposure. Results Using shotgun metagenome sequencing, participants before and after moderate-altitude exposure were classified into cluster BL (dominated by Blautia ) and cluster BA (dominated by Bacteroides ). Relative to cluster BL, cluster BA consisted predominantly of individuals exposed to moderate altitude. Compared to cluster BL, Cluster BA exhibited rewired metabolism of serum metabolites (i.e., amino acids, fatty acids and bile acids) and gut microbiota, lower inflammatory factor levels (i.e., tumor necrosis factor- α ( TNF-α )), and sparser correlations among these parameters. Individuals with baseline BL enterotype who transitioned to the BA enterotype following moderate-altitude exposure showed prominent improvement in fasting blood glucose (FBG) levels, with higher abundance of Bacteroidetes species (e.g., Bacteroides thetaiotaomicron , and Bacteroides uniformis ), but lower Proteobacteria species abundance (e.g., Escherichia coli ) and decreased L-Glutamic acid levels. Furthermore, fecal microbiota transplantation (FMT) from moderate-altitude exposed individuals to high-fat diet (HFD) fed mice confirmed increased Bacteroides abundance shifts associated with improvements in glucose homeostasis regulation and rewired amino acid metabolism. In addition, significant increases in alanine aminotransferase (ALT) levels but decreased serum creatinine (Scr), arterial oxygen saturation (SaO2), 4-Hydroxyproline, L-Glutamic acid, L-Asparagine, L-Threonine, L-Citrulline, L-Lysine and Isovaleric acid levels were identified as potentially important signals for individuals upon moderate altitude exposure, regardless of the gut microbiota enterotype. Conclusions Moderate altitude exposure could induce enterotype switching, and a Bacteroides -dominant enterotype may be a beneficial pattern of the gut microbiome related to host metabolism. Moderate-altitude exposure has potential implications for glycemic control, suggesting new avenues for managing FBG levels in future. Graphical abstract

Hamiltonian parameters estimation is crucial in condensed matter physics, but is time‐ and cost‐consuming. High‐resolution images provide detailed information of underlying physics, but extracting Hamiltonian parameters from them is difficult due to the huge Hilbert space. Here, a protocol for Hamiltonian parameters estimation from images based on a machine learning (ML) architecture is provided. It consists in learning a mapping between spin configurations and Hamiltonian parameters from a small amount of simulated images, applying the trained ML model to a single unexplored experimental image to estimate its key parameters, and predicting the corresponding materials properties by a physical model. The efficiency of the approach is demonstrated by reproducing the same spin configuration as the experimental one and predicting the coercive field, the saturation field, and even the volume of the experiment specimen accurately. The proposed approach paves a way to achieve a stable and efficient parameters estimation. Magnetic parameters can be efficiently estimated from an experimental observation of spin configuration by combining machine learning and micro‐magnetic simulation. The method includes learning a mapping from spin configurations to magnetic parameters on a small amount of micro‐magnetic simulated images and applying the trained machine learning model to a single unexplored experimental image to estimate its key parameters.

Aortic aneurysm/dissection (AAD) poses a life-threatening cardiovascular emergency with complex mechanisms and a notably high mortality rate. Zebrafish (Danio rerio) serve as valuable models for AAD due to the conservation of their three-layered arterial structure and genome with that of humans. However, the existing studies have predominantly focused on larval zebrafish, leaving a gap in our understanding of adult zebrafish. In this study, we utilized β-Aminopropionic Nitrile (BAPN) impregnation to induce AAD in both larval and adult zebrafish. Following induction, larval zebrafish exhibited a 28% widening of the dorsal aortic diameter (p < 0.0004, n = 10) and aortic arch malformations, with a high malformation rate of 75% (6/8). Conversely, adult zebrafish showed a 41.67% (5/12) mortality rate 22 days post-induction. At this time point, the dorsal aortic area had expanded by 2.46 times (p < 0.009), and the vessel wall demonstrated significant thickening (8.22 ± 2.23 μM vs. 26.38 ± 10.74 μM, p < 0.05). Pathological analysis revealed disruptions in the smooth muscle layer, contributing to a 58.33% aneurysm rate. Moreover, the expression levels of acta2, tagln, cnn1a, and cnn1b were decreased, indicating a weakened contractile phenotype. Transcriptome sequencing showed a significant overlap between the molecular features of zebrafish tissues post-BAPN treatment and those of AAD patients. Our findings present a straightforward and practical method for generating AAD models in both larval and adult zebrafish using BAPN.

Motion recognition, especially the distinction between high-speed and low-speed movements, is a challenging computational task that typically requires substantial resources. The extensive response range required to detect such variations in speed often exceeds the capabilities of traditional CMOS technology. This report introduces a SnS 2 -based in-sensor reservoir that offers an effective solution for detecting a variety of motion types at sensory terminals. By leveraging in-sensor reservoir computing, the device excels at classifying different motions across a wide velocity spectrum, providing a novel and promising method for motion recognition. The conductance of SnS 2 channel under light stimulation is governed by the trapping and recombination of photogenerated carriers at the inherent defect states, which contributes to the flexible optically dynamical sensing function of the device to varying illumination times. These attributes make the device versatile for both optical sensing and synaptic emulation. The findings suggest that such a SnS 2 -based device could be instrumental in advancing motion recognition capabilities for developing next-generation artificial intelligence systems.

The unprecedented development of Internet of Things results in the explosion of spatiotemporal signals generated by smart edge devices, leading to a surge of interest in real‐time learning of such data. This imposes a big challenge to conventional digital hardware because of physically separated memory and processing units and the transistor scaling limit. Memristors are deemed a solution for efficient and portable deep learning. However, their ionic resistive switching incurs large programming stochasticity and energy, compromising their advantages in real‐time learning spatiotemporal signals. To address the aforementioned issues, we propose a novel hardware–software codesign. Hardware‐wise, the stochasticity in memristor programming is leveraged to produce random matrices for efficient in‐memory computing. Software‐wise, random convolutional‐pooling architectures are integrated with echo‐state networks that compute with the hardware random matrices and make real‐time learning affordable. The synergy of the hardware and software not only improves the performance over conventional echo‐state networks, that is, 90.94% and 91.67% (compared to baselines 88.33% and 62.50%), but also retains 187.79× and 93.66× improvement of energy efficiency compared to the digital alternatives on the representative Human Activity Recognition Using Smartphones (HAR) and CRICKET datasets, respectively. These advantages make random convolutional echo‐state network (RCESN) a promising solution for the future smart edge hardware. Herein, a novel hardware–software codesign, the random memristor‐based convolutional echo‐state neural network (RCESN), is developed for edge spatiotemporal signal learning. Hardware‐wise, the stochasticity in memristor programming is leveraged to produce random matrices for efficient in‐memory computing. Software‐wise, random convolutional‐pooling architectures are paired with echo‐state networks that compute with the hardware random matrices and make real‐time learning affordable.

Exchange bias has increasingly practical significance in magnetoresistive and spintronic devices. However, the underlying mechanism of exchange bias in bulk compounds with the structural single‐phase and inhomogeneous magnetic phases is still elusive. Herein, based on experimental and simulation results, two important parameters are studied, i.e., the antiferromagnetic (AFM) volume fraction and the ferromagnetic (FM)/AFM interface area, which essentially determine the (spontaneous) exchange bias of Mn‐rich Ni44Co6Mn44‐xSn6+x (x = 0 ∼ 6) magnetic shape memory alloys. The substitution Sn for Mn changes magnetic ground state following the sequence of superparamagnetic/AFM → dilute spin glass/AFM → transition zone → cluster spin glass/AFM → FM, accompanying the growth of FM cluster and the weakening of AFM interactions. The results reveal that the magnetic ground state for exchange bias is optimized at transition zone between dilute spin glass/strong AFM and cluster spin glass/weak AFM, in which the optimal AFM volume fraction and FM/AFM interface area are achieved by tuning magnetic fields. A giant exchange bias field of 702.7 mT and a spontaneous exchange bias field of 318.7 mT are demonstrated. The work contributes to in‐depth understanding of (spontaneous) exchange bias in magnetically inhomogeneous compounds. Exchange bias has been investigated in Ni44Co6Mn44‐xSn6+x alloys. This work reveals that the AFM volume fraction and the FM/AFM interface area essentially determine (spontaneous) exchange bias of Mn‐rich Ni44Co6Mn44‐xSn6+x (x = 0 ∼ 6) magnetic shape memory alloys, which can be optimized by tuning the magnetic ground state (composition x) and magnetic field.

Autonomous one‐shot on‐the‐fly learning copes with the high privacy, small dataset, and in‐stream data at the edge. Implementing such learning on digital hardware suffers from the well‐known von‐Neumann and scaling bottlenecks. The optical neural networks featuring large parallelism, low latency, and high efficiency offer a promising solution. However, ex‐situ training of conventional optical networks, where optical path configuration and deep learning model optimization are separated, incurs hardware, energy and time overheads, and defeats the advantages in edge learning. Here, we introduced a bio‐inspired material‐algorithm co‐design to construct a hydrogel‐based optical Willshaw model (HOWM), manifesting Hebbian‐rule‐based structural plasticity for simultaneous optical path configuration and deep learning model optimization thanks to the underlying opto‐chemical reactions. We first employed the HOWM as an all optical in‐sensor AI processor for one‐shot pattern classification, association and denoising. We then leveraged HOWM to function as a ternary content addressable memory (TCAM) of an optical memory augmented neural network (MANN) for one‐shot learning the Omniglot dataset. The HOWM empowered one‐shot on‐the‐fly edge learning leads to 1000× boost of energy efficiency and 10× boost of speed, which paves the way for the next‐generation autonomous, efficient, and affordable smart edge systems. Biological synapses of human brain undergo a period of overproduction after birth, which is followed by consolidating part of the synapses and pruning the rest, bearing great significance to the development of intelligence as shown in Figure A. Such structural plasticity inspires a material‐algorithm co‐design, a hydrogel‐based optical Willshaw model (HOWM) in Figure B, thanks to the underlying opto‐chemical reactions of hydrogel shown in Figure C. The HOWM empowers one‐shot on‐the‐fly learning and leads to 1000× boost of energy efficiency and 10× boost of speed, which may pave the way for the next‐generation autonomous, efficient and affordable smart edge systems.

Aortic dissection (AD) represents a critical condition characterised by a tear in the inner lining of the aorta, leading to the leakage of blood into the layers of the aortic wall, posing a significant risk to life. However, the pathogenesis is unclear. In this study, scRNA‐seq was applied to cells derived from aortas of both AD and non‐AD donors (control) to unveil the cellular landscape. ScRNA‐seq data uncover significant cellular heterogeneity in AD aortas. Specifically, we observed an accumulation of CD4+ T cells, which contributed to inflammation and cell death, and abnormal collagen formation mediated by fibroblast cells in AD. Moreover, we revealed a greater prevalence of cell death, oxidative stress and senescence in AD aorta cells. Furthermore, we found a decrease in the percentage of vascular stem cells (VSCs), along with a repression in their ability to differentiate into contractile vascular smooth muscle cells (VSMCs). Finally, our data demonstrated that the PDGF signalling pathway was activated in AD. We found that PDGF activation could lead to VSMCs aberrant switch from contractile to synthetic phenotype, which could be ameliorated by PDGF inhibitor. This underscores the potential of the PDGF as a therapeutic target for AD. In summary, our study highlights the cellular heterogeneity and associated injuries within aortas affected by AD, including cell death, oxidative stress, senescence and dysregulation of signalling pathways influencing the aberrant phenotypic switch of VSMCs. These insights offer valuable contributions to understanding the molecular mechanisms underlying AD and present new avenues for therapeutic intervention in this condition.