Psychological studies on human subjects show that contrast detection learning promote learner＇s sensitivity to visual stimulus contrast. The underlying neural mechanisms remain unknown. In this study, three cats （Fe...Psychological studies on human subjects show that contrast detection learning promote learner＇s sensitivity to visual stimulus contrast. The underlying neural mechanisms remain unknown. In this study, three cats （Felis catus） were trained to perform monocularly a contrast detection task by two-altemative forced choice method. The perceptual ability of each cat improved remarkably with learning as indicated by a significantly increased contrast sensitivity to visual stimuli. The learning effect displayed an evident specificity to the eye employed for learning but could partially transfer to the naive eye, prompting the possibility that contrast detection learning might cause neural plasticity before and after the information from both eyes are merged in the visual pathway. Further, the contrast sensitivity improvement was evident basically around the spatial frequency （SF） used for learning, which suggested that contrast detection learning effect showed, to some extent, a SF specificity. This study indicates that cat exhibits a property of contrast detection learning similar to human subjects and can be used as an animal model for subsequent investigations on the neural correlates that mediate learning-induced contrast sensitivity improvement in humans.展开更多
Reliable anomaly detection in photovoltaic(PV)inverters is critical for ensuring operational efficiency and reducing maintenance costs in renewable energy systems.We introduce TRACE(Time series Representation learning...Reliable anomaly detection in photovoltaic(PV)inverters is critical for ensuring operational efficiency and reducing maintenance costs in renewable energy systems.We introduce TRACE(Time series Representation learning with Autoencoder-based Contrastive Embeddings),a self-supervised contrastive learning framework for multivariate time series anomaly detection in PV systems.TRACE employs a two-stage architecture:autoencoder-based representation learning with interchangeable backbones followed by contrastive training through a Siamese network.The framework generates semantically coherent augmentations by perturbing autoencoder reconstructions and applies three negative mining strategies to create challenging contrastive pairs.Comprehensive experiments on a real-world PV inverter dataset and two industrial benchmarks demonstrate TRACE’s superiority.Autoencoder-based augmentations deliver a 21.3%relative improvement in mean F1(0.616 vs.0.508)over traditional perturbation methods,with TransformerAE emerging as the optimal backbone architecture.While negative sampling strategies show dataset-specific advantages,their impact remains secondary to encoder capacity.TRACE with TransformerAE and reconstruction-error negatives consistently outperforms fourteen state-of-the-art time series anomaly detection methods,achieving highest F1 scores on all the three datasets while maintaining exceptional precision up to 0.99.Visualization analysis confirms TRACE’s capacity for early fault detection up to three days before failure and interpretable embedding separation.The framework addresses the fundamental challenge of label scarcity in industrial monitoring through self-supervised learning,providing a practical and transparent solution for predictive maintenance in PV systems and broader industrial applications.展开更多
Accurate recognition of low-contrast targets in complex visual environments is essential for advanced intelligent machine vision systems.Conventional photodetectors often suffer from a weak photoresponse and a linear ...Accurate recognition of low-contrast targets in complex visual environments is essential for advanced intelligent machine vision systems.Conventional photodetectors often suffer from a weak photoresponse and a linear dependence of photocurrent on light intensity,which restricts their ability to capture low-contrast features and makes them susceptible to noise.Inspired by the adaptive mechanisms of the human visual system,we present a molybdenum disulfide(MoS_(2))phototransistor with tunable sensitivity,in which the gate stack incorporates a heterostructure diode—composed of O-plasma-treated MoS_(2) and pristine MoS_(2)—that serves as the photosensitive layer.This configuration enables light-intensity-dependent modulation of the diode’s conductance,which dynamically in turn alters the voltage distribution across the gate dielectric and transistor channel,leading to a significant photoresponse.By modulating the gate voltage,the light response range can be finely tuned,maintaining high sensitivity to low-contrast targets while suppressing noise interference.Compared to conventional photodetectors,the proposed device achieves a 1000-fold improvement in sensitivity for low-contrast signal detection and exhibits significantly enhanced noise immunity.The intelligent machine vision system built on this device demonstrates exceptional performance in detecting low-contrast targets,underscoring its promise for next-generation machine vision applications.展开更多
Pulmonary arteriovenous fistula （PAVF） is a kind of malformation resulting in the abnormal vessels between pulmonary artery and pulmonary vein. Part of pulmonary arterial blood flows into pulmonary veins through the...Pulmonary arteriovenous fistula （PAVF） is a kind of malformation resulting in the abnormal vessels between pulmonary artery and pulmonary vein. Part of pulmonary arterial blood flows into pulmonary veins through the fistula and then arrives at left atrium, inducing the right-to-left shunt. Moreover, the emboli and bacteria can also flow directly through the PAVF into systemic circulation, which can cause thromboembolic diseases such as stroke.展开更多
基金Supported by Natural Science Foundation of Anhui Province(070413138)the foundation of Key Laboratory of Anhui Province and the Key Research Foundation from Education Department of Anhui Province(KJ2009A167)
文摘Psychological studies on human subjects show that contrast detection learning promote learner＇s sensitivity to visual stimulus contrast. The underlying neural mechanisms remain unknown. In this study, three cats （Felis catus） were trained to perform monocularly a contrast detection task by two-altemative forced choice method. The perceptual ability of each cat improved remarkably with learning as indicated by a significantly increased contrast sensitivity to visual stimuli. The learning effect displayed an evident specificity to the eye employed for learning but could partially transfer to the naive eye, prompting the possibility that contrast detection learning might cause neural plasticity before and after the information from both eyes are merged in the visual pathway. Further, the contrast sensitivity improvement was evident basically around the spatial frequency （SF） used for learning, which suggested that contrast detection learning effect showed, to some extent, a SF specificity. This study indicates that cat exhibits a property of contrast detection learning similar to human subjects and can be used as an animal model for subsequent investigations on the neural correlates that mediate learning-induced contrast sensitivity improvement in humans.
基金research project LongLife(020E 100583532),BMWE(German Federal Ministry for Eco-nomic Affairs and Energy),Germany.
文摘Reliable anomaly detection in photovoltaic(PV)inverters is critical for ensuring operational efficiency and reducing maintenance costs in renewable energy systems.We introduce TRACE(Time series Representation learning with Autoencoder-based Contrastive Embeddings),a self-supervised contrastive learning framework for multivariate time series anomaly detection in PV systems.TRACE employs a two-stage architecture:autoencoder-based representation learning with interchangeable backbones followed by contrastive training through a Siamese network.The framework generates semantically coherent augmentations by perturbing autoencoder reconstructions and applies three negative mining strategies to create challenging contrastive pairs.Comprehensive experiments on a real-world PV inverter dataset and two industrial benchmarks demonstrate TRACE’s superiority.Autoencoder-based augmentations deliver a 21.3%relative improvement in mean F1(0.616 vs.0.508)over traditional perturbation methods,with TransformerAE emerging as the optimal backbone architecture.While negative sampling strategies show dataset-specific advantages,their impact remains secondary to encoder capacity.TRACE with TransformerAE and reconstruction-error negatives consistently outperforms fourteen state-of-the-art time series anomaly detection methods,achieving highest F1 scores on all the three datasets while maintaining exceptional precision up to 0.99.Visualization analysis confirms TRACE’s capacity for early fault detection up to three days before failure and interpretable embedding separation.The framework addresses the fundamental challenge of label scarcity in industrial monitoring through self-supervised learning,providing a practical and transparent solution for predictive maintenance in PV systems and broader industrial applications.
基金supported by the National Key Research and Development Program of China(2021YFA1200801)the National Natural Science Foundation of China(No.62304226,52188101,62450124,62125406)+9 种基金the China Postdoctoral Science Foundation(2024T170946,2023M733574)the Excellent Youth Fund Project of Liaoning Province(2023JH3/10200003)the Outstanding Youth Fund Project of Liaoning Province(2025JH6/101100015)the Special Projects of the Central Government in Guidance of Local Science and Technology Development(2024010859-JH6/1006)the Special Research Assistantship Project of the Chinese Academy of Sciences(E455L502)the China Postdoctoral Science Foundation under Grant Number GZB20230776the Liaoning Provincial Key Laboratory of Public Opinion and Network Security Information System(d252453002)the Artificial Intelligence Technology Innovation Project of Liaoning Province(Grant No.2023JH26/10300019)the Young Top-notch Talents of the National High-level Talent Special Support Program,the basic scientific research project of universities funded by the Liaoning Provincial Department of Education(LJ212510140016)the Liaoning Province High-quality Industry-University Cooperation and Collaborative Education Project(241201160090747)。
文摘Accurate recognition of low-contrast targets in complex visual environments is essential for advanced intelligent machine vision systems.Conventional photodetectors often suffer from a weak photoresponse and a linear dependence of photocurrent on light intensity,which restricts their ability to capture low-contrast features and makes them susceptible to noise.Inspired by the adaptive mechanisms of the human visual system,we present a molybdenum disulfide(MoS_(2))phototransistor with tunable sensitivity,in which the gate stack incorporates a heterostructure diode—composed of O-plasma-treated MoS_(2) and pristine MoS_(2)—that serves as the photosensitive layer.This configuration enables light-intensity-dependent modulation of the diode’s conductance,which dynamically in turn alters the voltage distribution across the gate dielectric and transistor channel,leading to a significant photoresponse.By modulating the gate voltage,the light response range can be finely tuned,maintaining high sensitivity to low-contrast targets while suppressing noise interference.Compared to conventional photodetectors,the proposed device achieves a 1000-fold improvement in sensitivity for low-contrast signal detection and exhibits significantly enhanced noise immunity.The intelligent machine vision system built on this device demonstrates exceptional performance in detecting low-contrast targets,underscoring its promise for next-generation machine vision applications.
文摘Pulmonary arteriovenous fistula （PAVF） is a kind of malformation resulting in the abnormal vessels between pulmonary artery and pulmonary vein. Part of pulmonary arterial blood flows into pulmonary veins through the fistula and then arrives at left atrium, inducing the right-to-left shunt. Moreover, the emboli and bacteria can also flow directly through the PAVF into systemic circulation, which can cause thromboembolic diseases such as stroke.
