Graph Convolutional Neural Networks(GCNs)have been widely used in various fields due to their powerful capabilities in processing graph-structured data.However,GCNs encounter significant challenges when applied to sca...Graph Convolutional Neural Networks(GCNs)have been widely used in various fields due to their powerful capabilities in processing graph-structured data.However,GCNs encounter significant challenges when applied to scale-free graphs with power-law distributions,resulting in substantial distortions.Moreover,most of the existing GCN models are shallow structures,which restricts their ability to capture dependencies among distant nodes and more refined high-order node features in scale-free graphs with hierarchical structures.To more broadly and precisely apply GCNs to real-world graphs exhibiting scale-free or hierarchical structures and utilize multi-level aggregation of GCNs for capturing high-level information in local representations,we propose the Hyperbolic Deep Graph Convolutional Neural Network(HDGCNN),an end-to-end deep graph representation learning framework that can map scale-free graphs from Euclidean space to hyperbolic space.In HDGCNN,we define the fundamental operations of deep graph convolutional neural networks in hyperbolic space.Additionally,we introduce a hyperbolic feature transformation method based on identity mapping and a dense connection scheme based on a novel non-local message passing framework.In addition,we present a neighborhood aggregation method that combines initial structural featureswith hyperbolic attention coefficients.Through the above methods,HDGCNN effectively leverages both the structural features and node features of graph data,enabling enhanced exploration of non-local structural features and more refined node features in scale-free or hierarchical graphs.Experimental results demonstrate that HDGCNN achieves remarkable performance improvements over state-ofthe-art GCNs in node classification and link prediction tasks,even when utilizing low-dimensional embedding representations.Furthermore,when compared to shallow hyperbolic graph convolutional neural network models,HDGCNN exhibits notable advantages and performance enhancements.展开更多
This study assesses the suitability of convolutional neural networks(CNNs) for downscaling precipitation over East Africa in the context of seasonal forecasting. To achieve this, we design a set of experiments that co...This study assesses the suitability of convolutional neural networks(CNNs) for downscaling precipitation over East Africa in the context of seasonal forecasting. To achieve this, we design a set of experiments that compare different CNN configurations and deployed the best-performing architecture to downscale one-month lead seasonal forecasts of June–July–August–September(JJAS) precipitation from the Nanjing University of Information Science and Technology Climate Forecast System version 1.0(NUIST-CFS1.0) for 1982–2020. We also perform hyper-parameter optimization and introduce predictors over a larger area to include information about the main large-scale circulations that drive precipitation over the East Africa region, which improves the downscaling results. Finally, we validate the raw model and downscaled forecasts in terms of both deterministic and probabilistic verification metrics, as well as their ability to reproduce the observed precipitation extreme and spell indicator indices. The results show that the CNN-based downscaling consistently improves the raw model forecasts, with lower bias and more accurate representations of the observed mean and extreme precipitation spatial patterns. Besides, CNN-based downscaling yields a much more accurate forecast of extreme and spell indicators and reduces the significant relative biases exhibited by the raw model predictions. Moreover, our results show that CNN-based downscaling yields better skill scores than the raw model forecasts over most portions of East Africa. The results demonstrate the potential usefulness of CNN in downscaling seasonal precipitation predictions over East Africa,particularly in providing improved forecast products which are essential for end users.展开更多
Time series forecasting plays an important role in various fields, such as energy, finance, transport, and weather. Temporal convolutional networks (TCNs) based on dilated causal convolution have been widely used in t...Time series forecasting plays an important role in various fields, such as energy, finance, transport, and weather. Temporal convolutional networks (TCNs) based on dilated causal convolution have been widely used in time series forecasting. However, two problems weaken the performance of TCNs. One is that in dilated casual convolution, causal convolution leads to the receptive fields of outputs being concentrated in the earlier part of the input sequence, whereas the recent input information will be severely lost. The other is that the distribution shift problem in time series has not been adequately solved. To address the first problem, we propose a subsequence-based dilated convolution method (SDC). By using multiple convolutional filters to convolve elements of neighboring subsequences, the method extracts temporal features from a growing receptive field via a growing subsequence rather than a single element. Ultimately, the receptive field of each output element can cover the whole input sequence. To address the second problem, we propose a difference and compensation method (DCM). The method reduces the discrepancies between and within the input sequences by difference operations and then compensates the outputs for the information lost due to difference operations. Based on SDC and DCM, we further construct a temporal subsequence-based convolutional network with difference (TSCND) for time series forecasting. The experimental results show that TSCND can reduce prediction mean squared error by 7.3% and save runtime, compared with state-of-the-art models and vanilla TCN.展开更多
Geopolymer concrete emerges as a promising avenue for sustainable development and offers an effective solution to environmental problems.Its attributes as a non-toxic,low-carbon,and economical substitute for conventio...Geopolymer concrete emerges as a promising avenue for sustainable development and offers an effective solution to environmental problems.Its attributes as a non-toxic,low-carbon,and economical substitute for conventional cement concrete,coupled with its elevated compressive strength and reduced shrinkage properties,position it as a pivotal material for diverse applications spanning from architectural structures to transportation infrastructure.In this context,this study sets out the task of using machine learning(ML)algorithms to increase the accuracy and interpretability of predicting the compressive strength of geopolymer concrete in the civil engineering field.To achieve this goal,a new approach using convolutional neural networks(CNNs)has been adopted.This study focuses on creating a comprehensive dataset consisting of compositional and strength parameters of 162 geopolymer concrete mixes,all containing Class F fly ash.The selection of optimal input parameters is guided by two distinct criteria.The first criterion leverages insights garnered from previous research on the influence of individual features on compressive strength.The second criterion scrutinizes the impact of these features within the model’s predictive framework.Key to enhancing the CNN model’s performance is the meticulous determination of the optimal hyperparameters.Through a systematic trial-and-error process,the study ascertains the ideal number of epochs for data division and the optimal value of k for k-fold cross-validation—a technique vital to the model’s robustness.The model’s predictive prowess is rigorously assessed via a suite of performance metrics and comprehensive score analyses.Furthermore,the model’s adaptability is gauged by integrating a secondary dataset into its predictive framework,facilitating a comparative evaluation against conventional prediction methods.To unravel the intricacies of the CNN model’s learning trajectory,a loss plot is deployed to elucidate its learning rate.The study culminates in compelling findings that underscore the CNN model’s accurate prediction of geopolymer concrete compressive strength.To maximize the dataset’s potential,the application of bivariate plots unveils nuanced trends and interactions among variables,fortifying the consistency with earlier research.Evidenced by promising prediction accuracy,the study’s outcomes hold significant promise in guiding the development of innovative geopolymer concrete formulations,thereby reinforcing its role as an eco-conscious and robust construction material.The findings prove that the CNN model accurately estimated geopolymer concrete’s compressive strength.The results show that the prediction accuracy is promising and can be used for the development of new geopolymer concrete mixes.The outcomes not only underscore the significance of leveraging technology for sustainable construction practices but also pave the way for innovation and efficiency in the field of civil engineering.展开更多
Oscillation detection has been a hot research topic in industries due to the high incidence of oscillation loops and their negative impact on plant profitability.Although numerous automatic detection techniques have b...Oscillation detection has been a hot research topic in industries due to the high incidence of oscillation loops and their negative impact on plant profitability.Although numerous automatic detection techniques have been proposed,most of them can only address part of the practical difficulties.An oscillation is heuristically defined as a visually apparent periodic variation.However,manual visual inspection is labor-intensive and prone to missed detection.Convolutional neural networks(CNNs),inspired by animal visual systems,have been raised with powerful feature extraction capabilities.In this work,an exploration of the typical CNN models for visual oscillation detection is performed.Specifically,we tested MobileNet-V1,ShuffleNet-V2,Efficient Net-B0,and GhostNet models,and found that such a visual framework is well-suited for oscillation detection.The feasibility and validity of this framework are verified utilizing extensive numerical and industrial cases.Compared with state-of-theart oscillation detectors,the suggested framework is more straightforward and more robust to noise and mean-nonstationarity.In addition,this framework generalizes well and is capable of handling features that are not present in the training data,such as multiple oscillations and outliers.展开更多
In actual traffic scenarios,precise recognition of traffic participants,such as vehicles and pedestrians,is crucial for intelligent transportation.This study proposes an improved algorithm built on Mask-RCNN to enhanc...In actual traffic scenarios,precise recognition of traffic participants,such as vehicles and pedestrians,is crucial for intelligent transportation.This study proposes an improved algorithm built on Mask-RCNN to enhance the ability of autonomous driving systems to recognize traffic participants.The algorithmincorporates long and shortterm memory networks and the fused attention module(GSAM,GCT,and Spatial Attention Module)to enhance the algorithm’s capability to process both global and local information.Additionally,to increase the network’s initial operation stability,the original network activation function was replaced with Gaussian error linear unit.Experiments were conducted using the publicly available Cityscapes dataset.Comparing the test results,it was observed that the revised algorithmoutperformed the original algorithmin terms of AP_(50),AP_(75),and othermetrics by 8.7%and 9.6%for target detection and 12.5%and 13.3%for segmentation.展开更多
Recent advances in deep neural networks have shed new light on physics,engineering,and scientific computing.Reconciling the data-centered viewpoint with physical simulation is one of the research hotspots.The physicsi...Recent advances in deep neural networks have shed new light on physics,engineering,and scientific computing.Reconciling the data-centered viewpoint with physical simulation is one of the research hotspots.The physicsinformedneural network(PINN)is currently the most general framework,which is more popular due to theconvenience of constructing NNs and excellent generalization ability.The automatic differentiation(AD)-basedPINN model is suitable for the homogeneous scientific problem;however,it is unclear how AD can enforce fluxcontinuity across boundaries between cells of different properties where spatial heterogeneity is represented bygrid cells with different physical properties.In this work,we propose a criss-cross physics-informed convolutionalneural network(CC-PINN)learning architecture,aiming to learn the solution of parametric PDEs with spatialheterogeneity of physical properties.To achieve the seamless enforcement of flux continuity and integration ofphysicalmeaning into CNN,a predefined 2D convolutional layer is proposed to accurately express transmissibilitybetween adjacent cells.The efficacy of the proposedmethodwas evaluated through predictions of several petroleumreservoir problems with spatial heterogeneity and compared against state-of-the-art(PINN)through numericalanalysis as a benchmark,which demonstrated the superiority of the proposed method over the PINN.展开更多
Transfer learning could reduce the time and resources required by the training of new models and be therefore important for generalized applications of the trainedmachine learning algorithms.In this study,a transfer l...Transfer learning could reduce the time and resources required by the training of new models and be therefore important for generalized applications of the trainedmachine learning algorithms.In this study,a transfer learningenhanced convolutional neural network(CNN)was proposed to identify the gross weight and the axle weight of moving vehicles on the bridge.The proposed transfer learning-enhanced CNN model was expected to weigh different bridges based on a small amount of training datasets and provide high identification accuracy.First of all,a CNN algorithm for bridge weigh-in-motion(B-WIM)technology was proposed to identify the axle weight and the gross weight of the typical two-axle,three-axle,and five-axle vehicles as they crossed the bridge with different loading routes and speeds.Then,the pre-trained CNN model was transferred by fine-tuning to weigh themoving vehicle on another bridge.Finally,the identification accuracy and the amount of training data required were compared between the two CNN models.Results showed that the pre-trained CNN model using transfer learning for B-WIM technology could be successfully used for the identification of the axle weight and the gross weight for moving vehicles on another bridge while reducing the training data by 63%.Moreover,the recognition accuracy of the pre-trained CNN model using transfer learning was comparable to that of the original model,showing its promising potentials in the actual applications.展开更多
We design a new hybrid quantum-classical convolutional neural network(HQCCNN)model based on parameter quantum circuits.In this model,we use parameterized quantum circuits(PQCs)to redesign the convolutional layer in cl...We design a new hybrid quantum-classical convolutional neural network(HQCCNN)model based on parameter quantum circuits.In this model,we use parameterized quantum circuits(PQCs)to redesign the convolutional layer in classical convolutional neural networks,forming a new quantum convolutional layer to achieve unitary transformation of quantum states,enabling the model to more accurately extract hidden information from images.At the same time,we combine the classical fully connected layer with PQCs to form a new hybrid quantum-classical fully connected layer to further improve the accuracy of classification.Finally,we use the MNIST dataset to test the potential of the HQCCNN.The results indicate that the HQCCNN has good performance in solving classification problems.In binary classification tasks,the classification accuracy of numbers 5 and 7 is as high as 99.71%.In multivariate classification,the accuracy rate also reaches 98.51%.Finally,we compare the performance of the HQCCNN with other models and find that the HQCCNN has better classification performance and convergence speed.展开更多
Convolutional neural networks (CNNs) are widely used in image classification tasks, but their increasing model size and computation make them challenging to implement on embedded systems with constrained hardware reso...Convolutional neural networks (CNNs) are widely used in image classification tasks, but their increasing model size and computation make them challenging to implement on embedded systems with constrained hardware resources. To address this issue, the MobileNetV1 network was developed, which employs depthwise convolution to reduce network complexity. MobileNetV1 employs a stride of 2 in several convolutional layers to decrease the spatial resolution of feature maps, thereby lowering computational costs. However, this stride setting can lead to a loss of spatial information, particularly affecting the detection and representation of smaller objects or finer details in images. To maintain the trade-off between complexity and model performance, a lightweight convolutional neural network with hierarchical multi-scale feature fusion based on the MobileNetV1 network is proposed. The network consists of two main subnetworks. The first subnetwork uses a depthwise dilated separable convolution (DDSC) layer to learn imaging features with fewer parameters, which results in a lightweight and computationally inexpensive network. Furthermore, depthwise dilated convolution in DDSC layer effectively expands the field of view of filters, allowing them to incorporate a larger context. The second subnetwork is a hierarchical multi-scale feature fusion (HMFF) module that uses parallel multi-resolution branches architecture to process the input feature map in order to extract the multi-scale feature information of the input image. Experimental results on the CIFAR-10, Malaria, and KvasirV1 datasets demonstrate that the proposed method is efficient, reducing the network parameters and computational cost by 65.02% and 39.78%, respectively, while maintaining the network performance compared to the MobileNetV1 baseline.展开更多
Due to the structural dependencies among concurrent events in the knowledge graph and the substantial amount of sequential correlation information carried by temporally adjacent events,we propose an Independent Recurr...Due to the structural dependencies among concurrent events in the knowledge graph and the substantial amount of sequential correlation information carried by temporally adjacent events,we propose an Independent Recurrent Temporal Graph Convolution Networks(IndRT-GCNets)framework to efficiently and accurately capture event attribute information.The framework models the knowledge graph sequences to learn the evolutionary represen-tations of entities and relations within each period.Firstly,by utilizing the temporal graph convolution module in the evolutionary representation unit,the framework captures the structural dependency relationships within the knowledge graph in each period.Meanwhile,to achieve better event representation and establish effective correlations,an independent recurrent neural network is employed to implement auto-regressive modeling.Furthermore,static attributes of entities in the entity-relation events are constrained andmerged using a static graph constraint to obtain optimal entity representations.Finally,the evolution of entity and relation representations is utilized to predict events in the next subsequent step.On multiple real-world datasets such as Freebase13(FB13),Freebase 15k(FB15K),WordNet11(WN11),WordNet18(WN18),FB15K-237,WN18RR,YAGO3-10,and Nell-995,the results of multiple evaluation indicators show that our proposed IndRT-GCNets framework outperforms most existing models on knowledge reasoning tasks,which validates the effectiveness and robustness.展开更多
Generating realistic and synthetic video from text is a highly challenging task due to the multitude of issues involved,including digit deformation,noise interference between frames,blurred output,and the need for tem...Generating realistic and synthetic video from text is a highly challenging task due to the multitude of issues involved,including digit deformation,noise interference between frames,blurred output,and the need for temporal coherence across frames.In this paper,we propose a novel approach for generating coherent videos of moving digits from textual input using a Deep Deconvolutional Generative Adversarial Network(DD-GAN).The DDGAN comprises a Deep Deconvolutional Neural Network(DDNN)as a Generator(G)and a modified Deep Convolutional Neural Network(DCNN)as a Discriminator(D)to ensure temporal coherence between adjacent frames.The proposed research involves several steps.First,the input text is fed into a Long Short Term Memory(LSTM)based text encoder and then smoothed using Conditioning Augmentation(CA)techniques to enhance the effectiveness of the Generator(G).Next,using a DDNN to generate video frames by incorporating enhanced text and random noise and modifying a DCNN to act as a Discriminator(D),effectively distinguishing between generated and real videos.This research evaluates the quality of the generated videos using standard metrics like Inception Score(IS),Fréchet Inception Distance(FID),Fréchet Inception Distance for video(FID2vid),and Generative Adversarial Metric(GAM),along with a human study based on realism,coherence,and relevance.By conducting experiments on Single-Digit Bouncing MNIST GIFs(SBMG),Two-Digit Bouncing MNIST GIFs(TBMG),and a custom dataset of essential mathematics videos with related text,this research demonstrates significant improvements in both metrics and human study results,confirming the effectiveness of DD-GAN.This research also took the exciting challenge of generating preschool math videos from text,handling complex structures,digits,and symbols,and achieving successful results.The proposed research demonstrates promising results for generating coherent videos from textual input.展开更多
In intelligent perception and diagnosis of medical equipment,the visual and morphological changes in retinal vessels are closely related to the severity of cardiovascular diseases(e.g.,diabetes and hypertension).Intel...In intelligent perception and diagnosis of medical equipment,the visual and morphological changes in retinal vessels are closely related to the severity of cardiovascular diseases(e.g.,diabetes and hypertension).Intelligent auxiliary diagnosis of these diseases depends on the accuracy of the retinal vascular segmentation results.To address this challenge,we design a Dual-Branch-UNet framework,which comprises a Dual-Branch encoder structure for feature extraction based on the traditional U-Net model for medical image segmentation.To be more explicit,we utilize a novel parallel encoder made up of various convolutional modules to enhance the encoder portion of the original U-Net.Then,image features are combined at each layer to produce richer semantic data and the model’s capacity is adjusted to various input images.Meanwhile,in the lower sampling section,we give up pooling and conduct the lower sampling by convolution operation to control step size for information fusion.We also employ an attentionmodule in the decoder stage to filter the image noises so as to lessen the response of irrelevant features.Experiments are verified and compared on the DRIVE and ARIA datasets for retinal vessels segmentation.The proposed Dual-Branch-UNet has proved to be superior to other five typical state-of-the-art methods.展开更多
Problems:For people all over the world,cancer is one of the most feared diseases.Cancer is one of the major obstacles to improving life expectancy in countries around the world and one of the biggest causes of death b...Problems:For people all over the world,cancer is one of the most feared diseases.Cancer is one of the major obstacles to improving life expectancy in countries around the world and one of the biggest causes of death before the age of 70 in 112 countries.Among all kinds of cancers,breast cancer is the most common cancer for women.The data showed that female breast cancer had become one of themost common cancers.Aims:A large number of clinical trials have proved that if breast cancer is diagnosed at an early stage,it could give patients more treatment options and improve the treatment effect and survival ability.Based on this situation,there are many diagnostic methods for breast cancer,such as computer-aided diagnosis(CAD).Methods:We complete a comprehensive review of the diagnosis of breast cancer based on the convolutional neural network(CNN)after reviewing a sea of recent papers.Firstly,we introduce several different imaging modalities.The structure of CNN is given in the second part.After that,we introduce some public breast cancer data sets.Then,we divide the diagnosis of breast cancer into three different tasks:1.classification;2.detection;3.segmentation.Conclusion:Although this diagnosis with CNN has achieved great success,there are still some limitations.(i)There are too few good data sets.A good public breast cancer dataset needs to involve many aspects,such as professional medical knowledge,privacy issues,financial issues,dataset size,and so on.(ii)When the data set is too large,the CNN-based model needs a sea of computation and time to complete the diagnosis.(iii)It is easy to cause overfitting when using small data sets.展开更多
Imaging plates are widely used to detect alpha particles to track information,and the number of alpha particle tracks is affected by the overlapping and fading effects of the track information.In this study,an experim...Imaging plates are widely used to detect alpha particles to track information,and the number of alpha particle tracks is affected by the overlapping and fading effects of the track information.In this study,an experiment and a simulation were used to calibrate the efficiency parameter of an imaging plate,which was used to calculate the grayscale.Images were created by using grayscale,which trained the convolutional neural network to count the alpha tracks.The results demonstrated that the trained convolutional neural network can evaluate the alpha track counts based on the source and background images with a wider linear range,which was unaffected by the overlapping effect.The alpha track counts were unaffected by the fading effect within 60 min,where the calibrated formula for the fading effect was analyzed for 132.7 min.The detection efficiency of the trained convolutional neural network for inhomogeneous ^(241)Am sources(2π emission)was 0.6050±0.0399,whereas the efficiency curve of the photo-stimulated luminescence method was lower than that of the trained convolutional neural network.展开更多
Recent reactor antineutrino experiments have observed that the neutrino spectrum changes with the reactor core evolution and that the individual fissile isotope antineutrino spectra can be decomposed from the evolving...Recent reactor antineutrino experiments have observed that the neutrino spectrum changes with the reactor core evolution and that the individual fissile isotope antineutrino spectra can be decomposed from the evolving data,providing valuable information for the reactor model and data inconsistent problems.We propose a machine learning method by building a convolutional neural network based on a virtual experiment with a typical short-baseline reactor antineutrino experiment configuration:by utilizing the reactor evolution information,the major fissile isotope spectra are correctly extracted,and the uncertainties are evaluated using the Monte Carlo method.Validation tests show that the method is unbiased and introduces tiny extra uncertainties.展开更多
Neuromorphic photonic computing has emerged as a competitive computing paradigm to overcome the bottlenecks of the von-Neumann architecture.Linear weighting and nonlinear spike activation are two fundamental functions...Neuromorphic photonic computing has emerged as a competitive computing paradigm to overcome the bottlenecks of the von-Neumann architecture.Linear weighting and nonlinear spike activation are two fundamental functions of a photonic spiking neural network(PSNN).However,they are separately implemented with different photonic materials and devices,hindering the large-scale integration of PSNN.Here,we propose,fabricate and experimentally demonstrate a photonic neuro-synaptic chip enabling the simultaneous implementation of linear weighting and nonlinear spike activation based on a distributed feedback(DFB)laser with a saturable absorber(DFB-SA).A prototypical system is experimentally constructed to demonstrate the parallel weighted function and nonlinear spike activation.Furthermore,a fourchannel DFB-SA laser array is fabricated for realizing matrix convolution of a spiking convolutional neural network,achieving a recognition accuracy of 87%for the MNIST dataset.The fabricated neuro-synaptic chip offers a fundamental building block to construct the large-scale integrated PSNN chip.展开更多
A significant advantage of medical image processing is that it allows non-invasive exploration of internal anatomy in great detail.It is possible to create and study 3D models of anatomical structures to improve treatm...A significant advantage of medical image processing is that it allows non-invasive exploration of internal anatomy in great detail.It is possible to create and study 3D models of anatomical structures to improve treatment outcomes,develop more effective medical devices,or arrive at a more accurate diagnosis.This paper aims to present a fused evolutionary algorithm that takes advantage of both whale optimization and bacterial foraging optimization to optimize feature extraction.The classification process was conducted with the aid of a convolu-tional neural network(CNN)with dual graphs.Evaluation of the performance of the fused model is carried out with various methods.In the initial input Com-puter Tomography(CT)image,150 images are pre-processed and segmented to identify cancerous and non-cancerous nodules.The geometrical,statistical,struc-tural,and texture features are extracted from the preprocessed segmented image using various methods such as Gray-level co-occurrence matrix(GLCM),Histo-gram-oriented gradient features(HOG),and Gray-level dependence matrix(GLDM).To select the optimal features,a novel fusion approach known as Whale-Bacterial Foraging Optimization is proposed.For the classification of lung cancer,dual graph convolutional neural networks have been employed.A com-parison of classification algorithms and optimization algorithms has been con-ducted.According to the evaluated results,the proposed fused algorithm is successful with an accuracy of 98.72%in predicting lung tumors,and it outper-forms other conventional approaches.展开更多
At present,the acquisition of seismic data is developing toward high-precision and high-density methods.However,complex natural environments and cultural factors in many exploration areas cause difficulties in achievi...At present,the acquisition of seismic data is developing toward high-precision and high-density methods.However,complex natural environments and cultural factors in many exploration areas cause difficulties in achieving uniform and intensive acquisition,which makes complete seismic data collection impossible.Therefore,data reconstruction is required in the processing link to ensure imaging accuracy.Deep learning,as a new field in rapid development,presents clear advantages in feature extraction and modeling.In this study,the convolutional neural network deep learning algorithm is applied to seismic data reconstruction.Based on the convolutional neural network algorithm and combined with the characteristics of seismic data acquisition,two training strategies of supervised and unsupervised learning are designed to reconstruct sparse acquisition seismic records.First,a supervised learning strategy is proposed for labeled data,wherein the complete seismic data are segmented as the input of the training set and are randomly sampled before each training,thereby increasing the number of samples and the richness of features.Second,an unsupervised learning strategy based on large samples is proposed for unlabeled data,and the rolling segmentation method is used to update(pseudo)labels and training parameters in the training process.Through the reconstruction test of simulated and actual data,the deep learning algorithm based on a convolutional neural network shows better reconstruction quality and higher accuracy than compressed sensing based on Curvelet transform.展开更多
As deep learning techniques such as Convolutional Neural Networks(CNNs)are widely adopted,the complexity of CNNs is rapidly increasing due to the growing demand for CNN accelerator system-on-chip(SoC).Although convent...As deep learning techniques such as Convolutional Neural Networks(CNNs)are widely adopted,the complexity of CNNs is rapidly increasing due to the growing demand for CNN accelerator system-on-chip(SoC).Although conventional CNN accelerators can reduce the computational time of learning and inference tasks,they tend to occupy large chip areas due to many multiply-and-accumulate(MAC)operators when implemented in complex digital circuits,incurring excessive power consumption.To overcome these drawbacks,this work implements an analog convolutional filter consisting of an analog multiply-and-accumulate arithmetic circuit along with an analog-to-digital converter(ADC).This paper introduces the architecture of an analog convolutional kernel comprised of low-power ultra-small circuits for neural network accelerator chips.ADC is an essential component of the analog convolutional kernel used to convert the analog convolutional result to digital values to be stored in memory.This work presents the implementation of a highly low-power and area-efficient 12-bit Successive Approximation Register(SAR)ADC.Unlink most other SAR-ADCs with differential structure;the proposed ADC employs a single-ended capacitor array to support the preceding single-ended max-pooling circuit along with minimal power consumption.The SARADCimplementation also introduces a unique circuit that reduces kick-back noise to increase performance.It was implemented in a test chip using a 55 nm CMOS process.It demonstrates that the proposed ADC reduces Kick-back noise by 40%and consequently improves the ADC’s resolution by about 10%while providing a near rail-to-rail dynamic rangewith significantly lower power consumption than conventional ADCs.The ADC test chip shows a chip size of 4600μm^(2)with a power consumption of 6.6μW while providing an signal-to-noise-and-distortion ratio(SNDR)of 68.45 dB,corresponding to an effective number of bits(ENOB)of 11.07 bits.展开更多
基金supported by the National Natural Science Foundation of China-China State Railway Group Co.,Ltd.Railway Basic Research Joint Fund (Grant No.U2268217)the Scientific Funding for China Academy of Railway Sciences Corporation Limited (No.2021YJ183).
文摘Graph Convolutional Neural Networks(GCNs)have been widely used in various fields due to their powerful capabilities in processing graph-structured data.However,GCNs encounter significant challenges when applied to scale-free graphs with power-law distributions,resulting in substantial distortions.Moreover,most of the existing GCN models are shallow structures,which restricts their ability to capture dependencies among distant nodes and more refined high-order node features in scale-free graphs with hierarchical structures.To more broadly and precisely apply GCNs to real-world graphs exhibiting scale-free or hierarchical structures and utilize multi-level aggregation of GCNs for capturing high-level information in local representations,we propose the Hyperbolic Deep Graph Convolutional Neural Network(HDGCNN),an end-to-end deep graph representation learning framework that can map scale-free graphs from Euclidean space to hyperbolic space.In HDGCNN,we define the fundamental operations of deep graph convolutional neural networks in hyperbolic space.Additionally,we introduce a hyperbolic feature transformation method based on identity mapping and a dense connection scheme based on a novel non-local message passing framework.In addition,we present a neighborhood aggregation method that combines initial structural featureswith hyperbolic attention coefficients.Through the above methods,HDGCNN effectively leverages both the structural features and node features of graph data,enabling enhanced exploration of non-local structural features and more refined node features in scale-free or hierarchical graphs.Experimental results demonstrate that HDGCNN achieves remarkable performance improvements over state-ofthe-art GCNs in node classification and link prediction tasks,even when utilizing low-dimensional embedding representations.Furthermore,when compared to shallow hyperbolic graph convolutional neural network models,HDGCNN exhibits notable advantages and performance enhancements.
基金supported by the National Key Research and Development Program of China (Grant No.2020YFA0608000)the National Natural Science Foundation of China (Grant No. 42030605)the High-Performance Computing of Nanjing University of Information Science&Technology for their support of this work。
文摘This study assesses the suitability of convolutional neural networks(CNNs) for downscaling precipitation over East Africa in the context of seasonal forecasting. To achieve this, we design a set of experiments that compare different CNN configurations and deployed the best-performing architecture to downscale one-month lead seasonal forecasts of June–July–August–September(JJAS) precipitation from the Nanjing University of Information Science and Technology Climate Forecast System version 1.0(NUIST-CFS1.0) for 1982–2020. We also perform hyper-parameter optimization and introduce predictors over a larger area to include information about the main large-scale circulations that drive precipitation over the East Africa region, which improves the downscaling results. Finally, we validate the raw model and downscaled forecasts in terms of both deterministic and probabilistic verification metrics, as well as their ability to reproduce the observed precipitation extreme and spell indicator indices. The results show that the CNN-based downscaling consistently improves the raw model forecasts, with lower bias and more accurate representations of the observed mean and extreme precipitation spatial patterns. Besides, CNN-based downscaling yields a much more accurate forecast of extreme and spell indicators and reduces the significant relative biases exhibited by the raw model predictions. Moreover, our results show that CNN-based downscaling yields better skill scores than the raw model forecasts over most portions of East Africa. The results demonstrate the potential usefulness of CNN in downscaling seasonal precipitation predictions over East Africa,particularly in providing improved forecast products which are essential for end users.
基金supported by the National Key Research and Development Program of China(No.2018YFB2101300)the National Natural Science Foundation of China(Grant No.61871186)the Dean’s Fund of Engineering Research Center of Software/Hardware Co-Design Technology and Application,Ministry of Education(East China Normal University).
文摘Time series forecasting plays an important role in various fields, such as energy, finance, transport, and weather. Temporal convolutional networks (TCNs) based on dilated causal convolution have been widely used in time series forecasting. However, two problems weaken the performance of TCNs. One is that in dilated casual convolution, causal convolution leads to the receptive fields of outputs being concentrated in the earlier part of the input sequence, whereas the recent input information will be severely lost. The other is that the distribution shift problem in time series has not been adequately solved. To address the first problem, we propose a subsequence-based dilated convolution method (SDC). By using multiple convolutional filters to convolve elements of neighboring subsequences, the method extracts temporal features from a growing receptive field via a growing subsequence rather than a single element. Ultimately, the receptive field of each output element can cover the whole input sequence. To address the second problem, we propose a difference and compensation method (DCM). The method reduces the discrepancies between and within the input sequences by difference operations and then compensates the outputs for the information lost due to difference operations. Based on SDC and DCM, we further construct a temporal subsequence-based convolutional network with difference (TSCND) for time series forecasting. The experimental results show that TSCND can reduce prediction mean squared error by 7.3% and save runtime, compared with state-of-the-art models and vanilla TCN.
基金funded by the Researchers Supporting Program at King Saud University(RSPD2023R809).
文摘Geopolymer concrete emerges as a promising avenue for sustainable development and offers an effective solution to environmental problems.Its attributes as a non-toxic,low-carbon,and economical substitute for conventional cement concrete,coupled with its elevated compressive strength and reduced shrinkage properties,position it as a pivotal material for diverse applications spanning from architectural structures to transportation infrastructure.In this context,this study sets out the task of using machine learning(ML)algorithms to increase the accuracy and interpretability of predicting the compressive strength of geopolymer concrete in the civil engineering field.To achieve this goal,a new approach using convolutional neural networks(CNNs)has been adopted.This study focuses on creating a comprehensive dataset consisting of compositional and strength parameters of 162 geopolymer concrete mixes,all containing Class F fly ash.The selection of optimal input parameters is guided by two distinct criteria.The first criterion leverages insights garnered from previous research on the influence of individual features on compressive strength.The second criterion scrutinizes the impact of these features within the model’s predictive framework.Key to enhancing the CNN model’s performance is the meticulous determination of the optimal hyperparameters.Through a systematic trial-and-error process,the study ascertains the ideal number of epochs for data division and the optimal value of k for k-fold cross-validation—a technique vital to the model’s robustness.The model’s predictive prowess is rigorously assessed via a suite of performance metrics and comprehensive score analyses.Furthermore,the model’s adaptability is gauged by integrating a secondary dataset into its predictive framework,facilitating a comparative evaluation against conventional prediction methods.To unravel the intricacies of the CNN model’s learning trajectory,a loss plot is deployed to elucidate its learning rate.The study culminates in compelling findings that underscore the CNN model’s accurate prediction of geopolymer concrete compressive strength.To maximize the dataset’s potential,the application of bivariate plots unveils nuanced trends and interactions among variables,fortifying the consistency with earlier research.Evidenced by promising prediction accuracy,the study’s outcomes hold significant promise in guiding the development of innovative geopolymer concrete formulations,thereby reinforcing its role as an eco-conscious and robust construction material.The findings prove that the CNN model accurately estimated geopolymer concrete’s compressive strength.The results show that the prediction accuracy is promising and can be used for the development of new geopolymer concrete mixes.The outcomes not only underscore the significance of leveraging technology for sustainable construction practices but also pave the way for innovation and efficiency in the field of civil engineering.
基金the National Natural Science Foundation of China(62003298,62163036)the Major Project of Science and Technology of Yunnan Province(202202AD080005,202202AH080009)the Yunnan University Professional Degree Graduate Practice Innovation Fund Project(ZC-22222770)。
文摘Oscillation detection has been a hot research topic in industries due to the high incidence of oscillation loops and their negative impact on plant profitability.Although numerous automatic detection techniques have been proposed,most of them can only address part of the practical difficulties.An oscillation is heuristically defined as a visually apparent periodic variation.However,manual visual inspection is labor-intensive and prone to missed detection.Convolutional neural networks(CNNs),inspired by animal visual systems,have been raised with powerful feature extraction capabilities.In this work,an exploration of the typical CNN models for visual oscillation detection is performed.Specifically,we tested MobileNet-V1,ShuffleNet-V2,Efficient Net-B0,and GhostNet models,and found that such a visual framework is well-suited for oscillation detection.The feasibility and validity of this framework are verified utilizing extensive numerical and industrial cases.Compared with state-of-theart oscillation detectors,the suggested framework is more straightforward and more robust to noise and mean-nonstationarity.In addition,this framework generalizes well and is capable of handling features that are not present in the training data,such as multiple oscillations and outliers.
基金the National Natural Science Foundation of China(52175236)Qingdao People’s Livelihood Science and Technology Plan(19-6-1-88-nsh).
文摘In actual traffic scenarios,precise recognition of traffic participants,such as vehicles and pedestrians,is crucial for intelligent transportation.This study proposes an improved algorithm built on Mask-RCNN to enhance the ability of autonomous driving systems to recognize traffic participants.The algorithmincorporates long and shortterm memory networks and the fused attention module(GSAM,GCT,and Spatial Attention Module)to enhance the algorithm’s capability to process both global and local information.Additionally,to increase the network’s initial operation stability,the original network activation function was replaced with Gaussian error linear unit.Experiments were conducted using the publicly available Cityscapes dataset.Comparing the test results,it was observed that the revised algorithmoutperformed the original algorithmin terms of AP_(50),AP_(75),and othermetrics by 8.7%and 9.6%for target detection and 12.5%and 13.3%for segmentation.
基金the National Natural Science Foundation of China(No.52274048)Beijing Natural Science Foundation(No.3222037)+1 种基金the CNPC 14th Five-Year Perspective Fundamental Research Project(No.2021DJ2104)the Science Foundation of China University of Petroleum,Beijing(No.2462021YXZZ010).
文摘Recent advances in deep neural networks have shed new light on physics,engineering,and scientific computing.Reconciling the data-centered viewpoint with physical simulation is one of the research hotspots.The physicsinformedneural network(PINN)is currently the most general framework,which is more popular due to theconvenience of constructing NNs and excellent generalization ability.The automatic differentiation(AD)-basedPINN model is suitable for the homogeneous scientific problem;however,it is unclear how AD can enforce fluxcontinuity across boundaries between cells of different properties where spatial heterogeneity is represented bygrid cells with different physical properties.In this work,we propose a criss-cross physics-informed convolutionalneural network(CC-PINN)learning architecture,aiming to learn the solution of parametric PDEs with spatialheterogeneity of physical properties.To achieve the seamless enforcement of flux continuity and integration ofphysicalmeaning into CNN,a predefined 2D convolutional layer is proposed to accurately express transmissibilitybetween adjacent cells.The efficacy of the proposedmethodwas evaluated through predictions of several petroleumreservoir problems with spatial heterogeneity and compared against state-of-the-art(PINN)through numericalanalysis as a benchmark,which demonstrated the superiority of the proposed method over the PINN.
基金the financial support provided by the National Natural Science Foundation of China(Grant No.52208213)the Excellent Youth Foundation of Education Department in Hunan Province(Grant No.22B0141)+1 种基金the Xiaohe Sci-Tech Talents Special Funding under Hunan Provincial Sci-Tech Talents Sponsorship Program(2023TJ-X65)the Science Foundation of Xiangtan University(Grant No.21QDZ23).
文摘Transfer learning could reduce the time and resources required by the training of new models and be therefore important for generalized applications of the trainedmachine learning algorithms.In this study,a transfer learningenhanced convolutional neural network(CNN)was proposed to identify the gross weight and the axle weight of moving vehicles on the bridge.The proposed transfer learning-enhanced CNN model was expected to weigh different bridges based on a small amount of training datasets and provide high identification accuracy.First of all,a CNN algorithm for bridge weigh-in-motion(B-WIM)technology was proposed to identify the axle weight and the gross weight of the typical two-axle,three-axle,and five-axle vehicles as they crossed the bridge with different loading routes and speeds.Then,the pre-trained CNN model was transferred by fine-tuning to weigh themoving vehicle on another bridge.Finally,the identification accuracy and the amount of training data required were compared between the two CNN models.Results showed that the pre-trained CNN model using transfer learning for B-WIM technology could be successfully used for the identification of the axle weight and the gross weight for moving vehicles on another bridge while reducing the training data by 63%.Moreover,the recognition accuracy of the pre-trained CNN model using transfer learning was comparable to that of the original model,showing its promising potentials in the actual applications.
基金Project supported by the Natural Science Foundation of Shandong Province,China (Grant No.ZR2021MF049)the Joint Fund of Natural Science Foundation of Shandong Province (Grant Nos.ZR2022LLZ012 and ZR2021LLZ001)。
文摘We design a new hybrid quantum-classical convolutional neural network(HQCCNN)model based on parameter quantum circuits.In this model,we use parameterized quantum circuits(PQCs)to redesign the convolutional layer in classical convolutional neural networks,forming a new quantum convolutional layer to achieve unitary transformation of quantum states,enabling the model to more accurately extract hidden information from images.At the same time,we combine the classical fully connected layer with PQCs to form a new hybrid quantum-classical fully connected layer to further improve the accuracy of classification.Finally,we use the MNIST dataset to test the potential of the HQCCNN.The results indicate that the HQCCNN has good performance in solving classification problems.In binary classification tasks,the classification accuracy of numbers 5 and 7 is as high as 99.71%.In multivariate classification,the accuracy rate also reaches 98.51%.Finally,we compare the performance of the HQCCNN with other models and find that the HQCCNN has better classification performance and convergence speed.
文摘Convolutional neural networks (CNNs) are widely used in image classification tasks, but their increasing model size and computation make them challenging to implement on embedded systems with constrained hardware resources. To address this issue, the MobileNetV1 network was developed, which employs depthwise convolution to reduce network complexity. MobileNetV1 employs a stride of 2 in several convolutional layers to decrease the spatial resolution of feature maps, thereby lowering computational costs. However, this stride setting can lead to a loss of spatial information, particularly affecting the detection and representation of smaller objects or finer details in images. To maintain the trade-off between complexity and model performance, a lightweight convolutional neural network with hierarchical multi-scale feature fusion based on the MobileNetV1 network is proposed. The network consists of two main subnetworks. The first subnetwork uses a depthwise dilated separable convolution (DDSC) layer to learn imaging features with fewer parameters, which results in a lightweight and computationally inexpensive network. Furthermore, depthwise dilated convolution in DDSC layer effectively expands the field of view of filters, allowing them to incorporate a larger context. The second subnetwork is a hierarchical multi-scale feature fusion (HMFF) module that uses parallel multi-resolution branches architecture to process the input feature map in order to extract the multi-scale feature information of the input image. Experimental results on the CIFAR-10, Malaria, and KvasirV1 datasets demonstrate that the proposed method is efficient, reducing the network parameters and computational cost by 65.02% and 39.78%, respectively, while maintaining the network performance compared to the MobileNetV1 baseline.
基金the National Natural Science Founda-tion of China(62062062)hosted by Gulila Altenbek.
文摘Due to the structural dependencies among concurrent events in the knowledge graph and the substantial amount of sequential correlation information carried by temporally adjacent events,we propose an Independent Recurrent Temporal Graph Convolution Networks(IndRT-GCNets)framework to efficiently and accurately capture event attribute information.The framework models the knowledge graph sequences to learn the evolutionary represen-tations of entities and relations within each period.Firstly,by utilizing the temporal graph convolution module in the evolutionary representation unit,the framework captures the structural dependency relationships within the knowledge graph in each period.Meanwhile,to achieve better event representation and establish effective correlations,an independent recurrent neural network is employed to implement auto-regressive modeling.Furthermore,static attributes of entities in the entity-relation events are constrained andmerged using a static graph constraint to obtain optimal entity representations.Finally,the evolution of entity and relation representations is utilized to predict events in the next subsequent step.On multiple real-world datasets such as Freebase13(FB13),Freebase 15k(FB15K),WordNet11(WN11),WordNet18(WN18),FB15K-237,WN18RR,YAGO3-10,and Nell-995,the results of multiple evaluation indicators show that our proposed IndRT-GCNets framework outperforms most existing models on knowledge reasoning tasks,which validates the effectiveness and robustness.
基金supported by the General Program of the National Natural Science Foundation of China(Grant No.61977029).
文摘Generating realistic and synthetic video from text is a highly challenging task due to the multitude of issues involved,including digit deformation,noise interference between frames,blurred output,and the need for temporal coherence across frames.In this paper,we propose a novel approach for generating coherent videos of moving digits from textual input using a Deep Deconvolutional Generative Adversarial Network(DD-GAN).The DDGAN comprises a Deep Deconvolutional Neural Network(DDNN)as a Generator(G)and a modified Deep Convolutional Neural Network(DCNN)as a Discriminator(D)to ensure temporal coherence between adjacent frames.The proposed research involves several steps.First,the input text is fed into a Long Short Term Memory(LSTM)based text encoder and then smoothed using Conditioning Augmentation(CA)techniques to enhance the effectiveness of the Generator(G).Next,using a DDNN to generate video frames by incorporating enhanced text and random noise and modifying a DCNN to act as a Discriminator(D),effectively distinguishing between generated and real videos.This research evaluates the quality of the generated videos using standard metrics like Inception Score(IS),Fréchet Inception Distance(FID),Fréchet Inception Distance for video(FID2vid),and Generative Adversarial Metric(GAM),along with a human study based on realism,coherence,and relevance.By conducting experiments on Single-Digit Bouncing MNIST GIFs(SBMG),Two-Digit Bouncing MNIST GIFs(TBMG),and a custom dataset of essential mathematics videos with related text,this research demonstrates significant improvements in both metrics and human study results,confirming the effectiveness of DD-GAN.This research also took the exciting challenge of generating preschool math videos from text,handling complex structures,digits,and symbols,and achieving successful results.The proposed research demonstrates promising results for generating coherent videos from textual input.
基金supported by National Natural Science Foundation of China(NSFC)(61976123,62072213)Taishan Young Scholars Program of Shandong Provinceand Key Development Program for Basic Research of Shandong Province(ZR2020ZD44).
文摘In intelligent perception and diagnosis of medical equipment,the visual and morphological changes in retinal vessels are closely related to the severity of cardiovascular diseases(e.g.,diabetes and hypertension).Intelligent auxiliary diagnosis of these diseases depends on the accuracy of the retinal vascular segmentation results.To address this challenge,we design a Dual-Branch-UNet framework,which comprises a Dual-Branch encoder structure for feature extraction based on the traditional U-Net model for medical image segmentation.To be more explicit,we utilize a novel parallel encoder made up of various convolutional modules to enhance the encoder portion of the original U-Net.Then,image features are combined at each layer to produce richer semantic data and the model’s capacity is adjusted to various input images.Meanwhile,in the lower sampling section,we give up pooling and conduct the lower sampling by convolution operation to control step size for information fusion.We also employ an attentionmodule in the decoder stage to filter the image noises so as to lessen the response of irrelevant features.Experiments are verified and compared on the DRIVE and ARIA datasets for retinal vessels segmentation.The proposed Dual-Branch-UNet has proved to be superior to other five typical state-of-the-art methods.
基金supported by Medical Research Council Confidence in Concept Award,UK(MC_PC_17171)Royal Society International Exchanges Cost Share Award,UK(RP202G0230)+5 种基金BritishHeart Foundation Accelerator Award,UK(AA/18/3/34220)Hope Foundation for Cancer Research,UK(RM60G0680)Global Challenges Research Fund(GCRF),UK(P202PF11)Sino-UK Industrial Fund,UK(RP202G0289)LIAS Pioneering Partnerships Award,UK(P202ED10)Data Science Enhancement Fund,UK(P202RE237)。
文摘Problems:For people all over the world,cancer is one of the most feared diseases.Cancer is one of the major obstacles to improving life expectancy in countries around the world and one of the biggest causes of death before the age of 70 in 112 countries.Among all kinds of cancers,breast cancer is the most common cancer for women.The data showed that female breast cancer had become one of themost common cancers.Aims:A large number of clinical trials have proved that if breast cancer is diagnosed at an early stage,it could give patients more treatment options and improve the treatment effect and survival ability.Based on this situation,there are many diagnostic methods for breast cancer,such as computer-aided diagnosis(CAD).Methods:We complete a comprehensive review of the diagnosis of breast cancer based on the convolutional neural network(CNN)after reviewing a sea of recent papers.Firstly,we introduce several different imaging modalities.The structure of CNN is given in the second part.After that,we introduce some public breast cancer data sets.Then,we divide the diagnosis of breast cancer into three different tasks:1.classification;2.detection;3.segmentation.Conclusion:Although this diagnosis with CNN has achieved great success,there are still some limitations.(i)There are too few good data sets.A good public breast cancer dataset needs to involve many aspects,such as professional medical knowledge,privacy issues,financial issues,dataset size,and so on.(ii)When the data set is too large,the CNN-based model needs a sea of computation and time to complete the diagnosis.(iii)It is easy to cause overfitting when using small data sets.
基金supported by the Hunan Provincial Innovation Foundation for Postgraduates (No.QL20210228)the National Natural Science Foundation of China (No.12075112)the National Natural Science Foundation of China (No.12175102).
文摘Imaging plates are widely used to detect alpha particles to track information,and the number of alpha particle tracks is affected by the overlapping and fading effects of the track information.In this study,an experiment and a simulation were used to calibrate the efficiency parameter of an imaging plate,which was used to calculate the grayscale.Images were created by using grayscale,which trained the convolutional neural network to count the alpha tracks.The results demonstrated that the trained convolutional neural network can evaluate the alpha track counts based on the source and background images with a wider linear range,which was unaffected by the overlapping effect.The alpha track counts were unaffected by the fading effect within 60 min,where the calibrated formula for the fading effect was analyzed for 132.7 min.The detection efficiency of the trained convolutional neural network for inhomogeneous ^(241)Am sources(2π emission)was 0.6050±0.0399,whereas the efficiency curve of the photo-stimulated luminescence method was lower than that of the trained convolutional neural network.
基金supported by the National Natural Science Foundation of China (Nos.11675273 and 12075087)the Strategic Priority Research Program of the Chinese Academy of Sciences (No.XDA10011102)。
文摘Recent reactor antineutrino experiments have observed that the neutrino spectrum changes with the reactor core evolution and that the individual fissile isotope antineutrino spectra can be decomposed from the evolving data,providing valuable information for the reactor model and data inconsistent problems.We propose a machine learning method by building a convolutional neural network based on a virtual experiment with a typical short-baseline reactor antineutrino experiment configuration:by utilizing the reactor evolution information,the major fissile isotope spectra are correctly extracted,and the uncertainties are evaluated using the Monte Carlo method.Validation tests show that the method is unbiased and introduces tiny extra uncertainties.
基金financial supports from National Key Research and Development Program of China (2021YFB2801900,2021YFB2801901,2021YFB2801902,2021YFB2801904)National Natural Science Foundation of China (No.61974177)+1 种基金National Outstanding Youth Science Fund Project of National Natural Science Foundation of China (62022062)The Fundamental Research Funds for the Central Universities (QTZX23041).
文摘Neuromorphic photonic computing has emerged as a competitive computing paradigm to overcome the bottlenecks of the von-Neumann architecture.Linear weighting and nonlinear spike activation are two fundamental functions of a photonic spiking neural network(PSNN).However,they are separately implemented with different photonic materials and devices,hindering the large-scale integration of PSNN.Here,we propose,fabricate and experimentally demonstrate a photonic neuro-synaptic chip enabling the simultaneous implementation of linear weighting and nonlinear spike activation based on a distributed feedback(DFB)laser with a saturable absorber(DFB-SA).A prototypical system is experimentally constructed to demonstrate the parallel weighted function and nonlinear spike activation.Furthermore,a fourchannel DFB-SA laser array is fabricated for realizing matrix convolution of a spiking convolutional neural network,achieving a recognition accuracy of 87%for the MNIST dataset.The fabricated neuro-synaptic chip offers a fundamental building block to construct the large-scale integrated PSNN chip.
文摘A significant advantage of medical image processing is that it allows non-invasive exploration of internal anatomy in great detail.It is possible to create and study 3D models of anatomical structures to improve treatment outcomes,develop more effective medical devices,or arrive at a more accurate diagnosis.This paper aims to present a fused evolutionary algorithm that takes advantage of both whale optimization and bacterial foraging optimization to optimize feature extraction.The classification process was conducted with the aid of a convolu-tional neural network(CNN)with dual graphs.Evaluation of the performance of the fused model is carried out with various methods.In the initial input Com-puter Tomography(CT)image,150 images are pre-processed and segmented to identify cancerous and non-cancerous nodules.The geometrical,statistical,struc-tural,and texture features are extracted from the preprocessed segmented image using various methods such as Gray-level co-occurrence matrix(GLCM),Histo-gram-oriented gradient features(HOG),and Gray-level dependence matrix(GLDM).To select the optimal features,a novel fusion approach known as Whale-Bacterial Foraging Optimization is proposed.For the classification of lung cancer,dual graph convolutional neural networks have been employed.A com-parison of classification algorithms and optimization algorithms has been con-ducted.According to the evaluated results,the proposed fused algorithm is successful with an accuracy of 98.72%in predicting lung tumors,and it outper-forms other conventional approaches.
基金This study was supported by the National Natural Science Foundation of China under the project‘Research on the Dynamic Location of Receiver Points and Wave Field Separation Technology Based on Deep Learning in OBN Seismic Exploration’(No.42074140).
文摘At present,the acquisition of seismic data is developing toward high-precision and high-density methods.However,complex natural environments and cultural factors in many exploration areas cause difficulties in achieving uniform and intensive acquisition,which makes complete seismic data collection impossible.Therefore,data reconstruction is required in the processing link to ensure imaging accuracy.Deep learning,as a new field in rapid development,presents clear advantages in feature extraction and modeling.In this study,the convolutional neural network deep learning algorithm is applied to seismic data reconstruction.Based on the convolutional neural network algorithm and combined with the characteristics of seismic data acquisition,two training strategies of supervised and unsupervised learning are designed to reconstruct sparse acquisition seismic records.First,a supervised learning strategy is proposed for labeled data,wherein the complete seismic data are segmented as the input of the training set and are randomly sampled before each training,thereby increasing the number of samples and the richness of features.Second,an unsupervised learning strategy based on large samples is proposed for unlabeled data,and the rolling segmentation method is used to update(pseudo)labels and training parameters in the training process.Through the reconstruction test of simulated and actual data,the deep learning algorithm based on a convolutional neural network shows better reconstruction quality and higher accuracy than compressed sensing based on Curvelet transform.
基金This work was supported by the National Research Foundation of Korea(NRF)grant funded by theKorea government(MSIT)(No.2022R1A5A8026986)and supported by Institute of Information&communications Technology Planning&Evaluation(IITP)grant funded by the Korea government(MSIT)(No.2020-0-01304,Development of Self-learnable Mobile Recursive Neural Network Processor Technology)+3 种基金It was also supported by the MSIT(Ministry of Science and ICT),Korea,under the Grand Information Technology Research Center support program(IITP-2022-2020-0-01462)supervised by the“IITP(Institute for Information&communications Technology Planning&Evaluation)”supported by the National Research Foundation of Korea(NRF)grant funded by the Korea government(MSIT)(No.2021R1F1A1061314)In addition,this work was conducted during the research year of Chungbuk National University in 2020.
文摘As deep learning techniques such as Convolutional Neural Networks(CNNs)are widely adopted,the complexity of CNNs is rapidly increasing due to the growing demand for CNN accelerator system-on-chip(SoC).Although conventional CNN accelerators can reduce the computational time of learning and inference tasks,they tend to occupy large chip areas due to many multiply-and-accumulate(MAC)operators when implemented in complex digital circuits,incurring excessive power consumption.To overcome these drawbacks,this work implements an analog convolutional filter consisting of an analog multiply-and-accumulate arithmetic circuit along with an analog-to-digital converter(ADC).This paper introduces the architecture of an analog convolutional kernel comprised of low-power ultra-small circuits for neural network accelerator chips.ADC is an essential component of the analog convolutional kernel used to convert the analog convolutional result to digital values to be stored in memory.This work presents the implementation of a highly low-power and area-efficient 12-bit Successive Approximation Register(SAR)ADC.Unlink most other SAR-ADCs with differential structure;the proposed ADC employs a single-ended capacitor array to support the preceding single-ended max-pooling circuit along with minimal power consumption.The SARADCimplementation also introduces a unique circuit that reduces kick-back noise to increase performance.It was implemented in a test chip using a 55 nm CMOS process.It demonstrates that the proposed ADC reduces Kick-back noise by 40%and consequently improves the ADC’s resolution by about 10%while providing a near rail-to-rail dynamic rangewith significantly lower power consumption than conventional ADCs.The ADC test chip shows a chip size of 4600μm^(2)with a power consumption of 6.6μW while providing an signal-to-noise-and-distortion ratio(SNDR)of 68.45 dB,corresponding to an effective number of bits(ENOB)of 11.07 bits.
