Deep learning(DL)models have demonstrated significant value in computational perception,superresolution imaging,ultra-precision measurement,and photonic device design.In optical communication signal recognition,DL mod...Deep learning(DL)models have demonstrated significant value in computational perception,superresolution imaging,ultra-precision measurement,and photonic device design.In optical communication signal recognition,DL models can achieve fast and accurate identification.However,in high-capacity optical communication systems represented by orbital angular momentum(OAM)beams,neural networks often suffer from excessive parameter sizes and demand large training datasets.To address these challenges,we report a lightweight MobileNetV1 model optimized with efficient channel attention to perform OAM mode recognition after transmission through free space and underwater tank environments.Experimental results show that in simulated small-sample classification tasks with five samples per class,the proposed model achieves an accuracy of 99.67%even under moderate turbulence conditions,outperforming four other DL models.In addition,for experimental datasets collected from both atmospheric turbulence and underwater environments,the model consistently achieves recognition accuracies exceeding 90%,demonstrating strong generalization ability and a 77%reduction in parameter count compared to traditional convolutional neural network(CNN)-based DL models.We provide a new approach for deploying lightweight DL algorithms on resource-constrained embedded optical signal detection devices.展开更多
By defining fuzzy valued simple functions and giving L1(μ) approximations of fuzzy valued integrably bounded functions by such simple functions, the paper analyses by L1(μ)-norm the approximation capability of four-...By defining fuzzy valued simple functions and giving L1(μ) approximations of fuzzy valued integrably bounded functions by such simple functions, the paper analyses by L1(μ)-norm the approximation capability of four-layer feedforward regular fuzzy neural networks to the fuzzy valued integrably bounded function F : Rn → FcO(R). That is, if the transfer functionσ: R→R is non-polynomial and integrable function on each finite interval, F may be innorm approximated by fuzzy valued functions defined as to anydegree of accuracy. Finally some real examples demonstrate the conclusions.展开更多
Micro-expressions(ME)recognition is a complex task that requires advanced techniques to extract informative features fromfacial expressions.Numerous deep neural networks(DNNs)with convolutional structures have been pr...Micro-expressions(ME)recognition is a complex task that requires advanced techniques to extract informative features fromfacial expressions.Numerous deep neural networks(DNNs)with convolutional structures have been proposed.However,unlike DNNs,shallow convolutional neural networks often outperform deeper models in mitigating overfitting,particularly with small datasets.Still,many of these methods rely on a single feature for recognition,resulting in an insufficient ability to extract highly effective features.To address this limitation,in this paper,an Improved Dual-stream Shallow Convolutional Neural Network based on an Extreme Gradient Boosting Algorithm(IDSSCNN-XgBoost)is introduced for ME Recognition.The proposed method utilizes a dual-stream architecture where motion vectors(temporal features)are extracted using Optical Flow TV-L1 and amplify subtle changes(spatial features)via EulerianVideoMagnification(EVM).These features are processed by IDSSCNN,with an attention mechanism applied to refine the extracted effective features.The outputs are then fused,concatenated,and classified using the XgBoost algorithm.This comprehensive approach significantly improves recognition accuracy by leveraging the strengths of both temporal and spatial information,supported by the robust classification power of XgBoost.The proposed method is evaluated on three publicly available ME databases named Chinese Academy of Sciences Micro-expression Database(CASMEII),Spontaneous Micro-Expression Database(SMICHS),and Spontaneous Actions and Micro-Movements(SAMM).Experimental results indicate that the proposed model can achieve outstanding results compared to recent models.The accuracy results are 79.01%,69.22%,and 68.99%on CASMEII,SMIC-HS,and SAMM,and the F1-score are 75.47%,68.91%,and 63.84%,respectively.The proposed method has the advantage of operational efficiency and less computational time.展开更多
With a spatial resolution of 50 m,a revisit time of three days,and a swath of 950 km,the coastal zone imager(CZI)offers great potential in monitoring coastal zone dynamics.Accurate atmo-spheric correction(AC)is needed...With a spatial resolution of 50 m,a revisit time of three days,and a swath of 950 km,the coastal zone imager(CZI)offers great potential in monitoring coastal zone dynamics.Accurate atmo-spheric correction(AC)is needed to exploit the potential of quantitative ocean color inversion.However,due to the band setting of CZI,the AC over coastal waters in the western Pacific region with complex optical properties cannot be realized easily.This research introduces a novel neural network(NN)AC algorithm for CZI data over coastal waters.Total 100,000 match-ups of HY-1 C CZI-observed reflectance at the top-of-atmosphere and Operational Land Imager(OLI)-retrieved high-quality remote sensing reflectance(Rrs)at the CZI bands are built to train the NN model.These reflectance data are obtained from the standard AC algorithm in the SeaDAS.Results indicate that the distributions of the CZI retrieved Rrs were consistent with the quasi-synchronous OLI data,but the spatial information from the CZI is more detailed.Then,the accuracy of the CZI data for AC is evaluated using the multi-source in-situ data.Results further show that the NN-AC can successfully retrieve Rrs for CZI and the coefficients of determination in the blue,green,red,and near-infrared bands were 0.70,0.77,0.76,and 0.67,respectively.The NN algorithm does not depend on shortwave-infrared bands and runs very fast once properly trained.展开更多
0-1 programming is a special case of the integer programming, which is commonly encountered in many optimization problems. Neural network and its general energy function are presented for 0-1 optimization problem. The...0-1 programming is a special case of the integer programming, which is commonly encountered in many optimization problems. Neural network and its general energy function are presented for 0-1 optimization problem. Then, the 0-1 optimization problems are solved by a neural network model with transient chaotic dynamics (TCNN). Numerical simulations of two typical 0-1 optimization problems show that TCNN can overcome HNN's main drawbacks that it suffers from the local minimum and can search for the global optimal solutions in to solveing 0-1 optimization problems.展开更多
A neural network with a feed forward topology and Bayesian regularization training algorithm is used to predict the austenite formation temperatures （At1 and A13） by considering the percentage of alloying elements i...A neural network with a feed forward topology and Bayesian regularization training algorithm is used to predict the austenite formation temperatures （At1 and A13） by considering the percentage of alloying elements in chemical composition of steel. The data base used here involves a large variety of different steel types such as struc- tural steels, stainless steels, rail steels, spring steels, high temperature creep resisting steels and tool steels. Scatter diagrams and mean relative error （MRE） statistical criteria are used to compare the performance of developed neural network with the results of Andrew% empirical equations and a feed forward neural network with ＂gradient descent with momentum＂ training algorithm. The results showed that Bayesian regularization neural network has the best performance. Also, due to the satisfactory results of the developed neural network, it was used to investigate the effect of the chemical composition on Ac1 and At3 temperatures. Results are in accordance with materials science theories.展开更多
The main task of this work is to design a control system for a small tail-sitter Unmanned Aerial Vehicle(UAV)during the transition process.Although reasonable control performance can be obtained through a well-tuned s...The main task of this work is to design a control system for a small tail-sitter Unmanned Aerial Vehicle(UAV)during the transition process.Although reasonable control performance can be obtained through a well-tuned single PID or cascade PID control architecture under nominal conditions,large or fast time-varying disturbances and a wide range of changes in the equilibrium point bring nonlinear characteristics to the transition control during the transition process,which leads to control precision degradation.Meanwhile,the PID controller’s tuning method relies on engineering experiences to a certain extent and the controller parameters need to be retuned under different working conditions,which limits the rapid deployment and preliminary validation.Based on the above issues,a novel control architecture of L1 neural network adaptive control associated with PID control is proposed to improve the compensation ability during the transition process and guarantee the security transition.The L1 neural network adaptive control is revised to solve the multi-input and multi-output problem of the tail-sitter UAV system in this study.Finally,the transition characteristics of the time setting difference between the desired transition speed and the desired transition pitch angle are analyzed.展开更多
Integrated with sensors,processors,and radio frequency(RF)communication modules,intelligent bearing could achieve the autonomous perception and autonomous decision-making,guarantying the safety and reliability during ...Integrated with sensors,processors,and radio frequency(RF)communication modules,intelligent bearing could achieve the autonomous perception and autonomous decision-making,guarantying the safety and reliability during their use.However,because of the resource limitations of the end device,processors in the intelligent bearing are unable to carry the computational load of deep learning models like convolutional neural network(CNN),which involves a great amount of multiplicative operations.To minimize the computation cost of the conventional CNN,based on the idea of AdderNet,a 1-D adder neural network with a wide first-layer kernel(WAddNN)suitable for bearing fault diagnosis is proposed in this paper.The proposed method uses the l1-norm distance between filters and input features as the output response,thus making the whole network almost free of multiplicative operations.The whole model takes the original signal as the input,uses a wide kernel in the first adder layer to extract features and suppress the high frequency noise,and then uses two layers of small kernels for nonlinear mapping.Through experimental comparison with CNN models of the same structure,WAddNN is able to achieve a similar accuracy as CNN models with significantly reduced computational cost.The proposed model provides a new fault diagnosis method for intelligent bearings with limited resources.展开更多
Due to the increasing variety of information and services carried by optical networks, the survivability of network becomes an important problem in current research. The fault location of OTN is of great significance ...Due to the increasing variety of information and services carried by optical networks, the survivability of network becomes an important problem in current research. The fault location of OTN is of great significance for studying the survivability of optical networks. Firstly, a three-channel network model is established and analyzing common alarm data, the fault monitoring points and common fault points are carried out. The artificial neural network is introduced into the fault location field of OTN and it is used to judge whether the possible fault point exists or not. But one of the obvious limitations of general neural networks is that they receive a fixedsize vector as input and produce a fixed-size vector as the output. Not only that, these models is even fixed for mapping operations (for example, the number of layers in the model). The difference between the recurrent neural network and general neural networks is that it can operate on the sequence. In spite of the fact that the gradient disappears and the gradient explodes still exist in the neural network, the method of gradient shearing or weight regularization is adopted to solve this problem, and choose the LSTM (long-short term memory networks) to locate the fault. The output uses the concept of membership degree of fuzzy theory to express the possible fault point with the probability from 0 to 1. Priority is given to the treatment of fault points with high probability. The concept of F-Measure is also introduced, and the positioning effect is measured by using location time, MSE and F-Measure. The experiment shows that both LSTM and BP neural network can locate the fault of optical transport network well, but the overall effect of LSTM is better. The localization time of LSTM is shorter than that of BP neural network, and the F1-score of LSTM can reach 0.961566888396156 after 45 iterations, which meets the accuracy and real-time requirements of fault location. Therefore, it has good application prospect and practical value to introduce neural network into the fault location field of optical transport network.展开更多
In order to directly construct the mapping between multiple state parameters and remaining useful life(RUL),and reduce the interference of random error on prediction accuracy,a RUL prediction model of aeroengine based...In order to directly construct the mapping between multiple state parameters and remaining useful life(RUL),and reduce the interference of random error on prediction accuracy,a RUL prediction model of aeroengine based on principal component analysis(PCA)and one-dimensional convolution neural network(1D-CNN)is proposed in this paper.Firstly,multiple state parameters corresponding to massive cycles of aeroengine are collected and brought into PCA for dimensionality reduction,and principal components are extracted for further time series prediction.Secondly,the 1D-CNN model is constructed to directly study the mapping between principal components and RUL.Multiple convolution and pooling operations are applied for deep feature extraction,and the end-to-end RUL prediction of aeroengine can be realized.Experimental results show that the most effective principal component from the multiple state parameters can be obtained by PCA,and the long time series of multiple state parameters can be directly mapped to RUL by 1D-CNN,so as to improve the efficiency and accuracy of RUL prediction.Compared with other traditional models,the proposed method also has lower prediction error and better robustness.展开更多
Neural stem cells(NSCs) and imprinted genes play an important role in brain development. On historical grounds, these two determinants have been largely studied independently of each other. Recent evidence suggests, h...Neural stem cells(NSCs) and imprinted genes play an important role in brain development. On historical grounds, these two determinants have been largely studied independently of each other. Recent evidence suggests, however, that NSCs can reset select genomic imprints to prevent precocious depletion of the stem cell reservoir. Moreover, imprinted genes like the transcriptional regulator Zac1 can fine tune neuronal vs astroglial differentiation of NSCs. Zac1 binds in a sequence-specific manner to pro-neuronal and imprinted genes to confer transcriptional regulation and furthermore coregulates members of the p53-family in NSCs. At the genome scale, Zac1 is a central hub of an imprinted gene network comprising genes with animportant role for NSC quiescence, proliferation and differentiation. Overall, transcriptional, epigenomic, and genomic mechanisms seem to coordinate the functional relationships of NSCs and imprinted genes from development to maturation, and possibly aging.展开更多
The hot deformation behavior of TI (18W-4Cr-1V) high-speed steel was investigated by means of continuous compression tests performed on Gleeble 1500 thermomechan- ical simulator in a wide range of tempemtures (950℃...The hot deformation behavior of TI (18W-4Cr-1V) high-speed steel was investigated by means of continuous compression tests performed on Gleeble 1500 thermomechan- ical simulator in a wide range of tempemtures (950℃-1150℃) with strain rotes of 0.001s-1-10s-1 and true strains of 0-0. 7. The flow stress at the above hot defor- mation conditions is predicted by using BP artificial neural network. The architecture of network includes there are three input parameters:strain rate,temperature T and true strain , and just one output parameter, the flow stress ,2 hidden layers are adopted, the first hidden layer includes 9 neurons and second 10 negroes. It has been verified that BP artificial neural network with 3-9-10-1 architecture can predict flow stress of high-speed steel during hot deformation very well. Compared with the prediction method of flow stress by using Zaped-Holloman parumeter and hyperbolic sine stress function, the prediction method by using BP artificial neurul network has higher efficiency and accuracy.展开更多
In developing Artificial Neural Networks(ANNs),the available dataset is split into three categories:training,validation and testing.However,an important problem arises:How to trust the predic-tion provided by a partic...In developing Artificial Neural Networks(ANNs),the available dataset is split into three categories:training,validation and testing.However,an important problem arises:How to trust the predic-tion provided by a particular ANN?Due to the randomness related to the network itself(architecture,initialization and learning procedure),there is usually no best choice.Considering this issue,we provide a framework,which captures the randomness related to the network itself.The idea is to perform several training and test trials based on the Jackknife resampling method.Jackknife consists of iteratively deleting a single observation each time from the sample and recomputing the ANN on the rest of the sample data.Consequently,interval prediction is available instead of point prediction.The proposed method was applied and tested using pH,Ca and P data obtained by analyzing 118 georeferenced soil points.The results,based on the dataset size simulation,showed that 60%reduction in available dataset offers compatible accuracy in relation to full dataset,and therefore a higher cost of sampling in the field would not be necessary.The re-sampling method spatially characterizes the points of greater or lesser accuracy and uncertainty.The re-sampling method increased the success rate by using interval prediction instead of using the mean as the most probable value.Although we restrict it to the regression neural network model,the resampling method proposed can also be extended to other modern statistical tools,such as Kriging,Least Squares Collocation(LSC),Convolutional Neural Network(CNN),and so on.展开更多
The geophysical model function （GMF） describes the relationship between backscattering and sea surface wind, so that wind vec- tors can be retrieved from backscattering measurement. The GMF plays an important role i...The geophysical model function （GMF） describes the relationship between backscattering and sea surface wind, so that wind vec- tors can be retrieved from backscattering measurement. The GMF plays an important role in ocean wind vector retrievals, its performance will directly influence the accuracy of the retrieved wind vector. Neural network （NN） approach is used to develop a unified GMF for C-band and Ku-band （NN-GMF）. Empirical GMF CMOIM and QSCAT-1 are used to generate the simulated training data-set, and Gaussian noise at a signal noise ratio of 30 dB is added to the data-set to simulate the noise in the backscat- tering measurement. The NN-GMF employs radio frequency as an additional parameter, so it can be applied for both C-band and Ku-band. Analyses show that the %predicted by the NN-GMF is comparable with the σpredicted by CMOIM and QSCAT-1. Also the wind vectors retrieved from the NN-GMF and empirical GMF CMOIM and QSCAT-1 are comparable, indicating that the NN-GMF is as effective as the empirical GMF, and has the advantages of the universal form.展开更多
Audiovisual speech recognition is an emerging research topic.Lipreading is the recognition of what someone is saying using visual information,primarily lip movements.In this study,we created a custom dataset for India...Audiovisual speech recognition is an emerging research topic.Lipreading is the recognition of what someone is saying using visual information,primarily lip movements.In this study,we created a custom dataset for Indian English linguistics and categorized it into three main categories:(1)audio recognition,(2)visual feature extraction,and(3)combined audio and visual recognition.Audio features were extracted using the mel-frequency cepstral coefficient,and classification was performed using a one-dimension convolutional neural network.Visual feature extraction uses Dlib and then classifies visual speech using a long short-term memory type of recurrent neural networks.Finally,integration was performed using a deep convolutional network.The audio speech of Indian English was successfully recognized with accuracies of 93.67%and 91.53%,respectively,using testing data from 200 epochs.The training accuracy for visual speech recognition using the Indian English dataset was 77.48%and the test accuracy was 76.19%using 60 epochs.After integration,the accuracies of audiovisual speech recognition using the Indian English dataset for training and testing were 94.67%and 91.75%,respectively.展开更多
Memristor-based neuromorphic computing shows great potential for high-speed and high-throughput signal processing applications,such as electroencephalogram(EEG)signal processing.Nonetheless,the size of one-transistor ...Memristor-based neuromorphic computing shows great potential for high-speed and high-throughput signal processing applications,such as electroencephalogram(EEG)signal processing.Nonetheless,the size of one-transistor one-resistor(1T1R)memristor arrays is limited by the non-ideality of the devices,which prevents the hardware implementation of large and complex networks.In this work,we propose the depthwise separable convolution and bidirectional gate recurrent unit(DSC-BiGRU)network,a lightweight and highly robust hybrid neural network based on 1T1R arrays that enables efficient processing of EEG signals in the temporal,frequency and spatial domains by hybridizing DSC and BiGRU blocks.The network size is reduced and the network robustness is improved while ensuring the network classification accuracy.In the simulation,the measured non-idealities of the 1T1R array are brought into the network through statistical analysis.Compared with traditional convolutional networks,the network parameters are reduced by 95%and the network classification accuracy is improved by 21%at a 95%array yield rate and 5%tolerable error.This work demonstrates that lightweight and highly robust networks based on memristor arrays hold great promise for applications that rely on low consumption and high efficiency.展开更多
With the rapid growth of the Internet in recent years, the ability to analyze and identify its users has become increasingly important. Authorship analysis provides a means to glean information about the author of a d...With the rapid growth of the Internet in recent years, the ability to analyze and identify its users has become increasingly important. Authorship analysis provides a means to glean information about the author of a document originating from the internet or elsewhere, including but not limited to the author’s gender. There are well-known linguistic differences between the writing of men and women, and these differences can be effectively used to predict the gender of a document’s author. Capitalizing on these linguistic nuances, this study uses a set of stylometric features and a set of word count features to facilitate automatic gender discrimination on emails from the popular Enron email dataset. These features are used in conjunction with the Modified Balanced Winnow Neural Network proposed by Carvalho and Cohen, an improvement on the original Balanced Winnow created by Littlestone. Experiments with the Modified Balanced Winnow show that it is effectively able to discriminate gender using both stylometric and word count features, with the word count features providing superior results.展开更多
基金supported by the China Postdoctoral Science Foundation(Grant No.2024M760415)the Natural Science Foundation of Guangdong Province(Grant No.2022A1515140118).
文摘Deep learning(DL)models have demonstrated significant value in computational perception,superresolution imaging,ultra-precision measurement,and photonic device design.In optical communication signal recognition,DL models can achieve fast and accurate identification.However,in high-capacity optical communication systems represented by orbital angular momentum(OAM)beams,neural networks often suffer from excessive parameter sizes and demand large training datasets.To address these challenges,we report a lightweight MobileNetV1 model optimized with efficient channel attention to perform OAM mode recognition after transmission through free space and underwater tank environments.Experimental results show that in simulated small-sample classification tasks with five samples per class,the proposed model achieves an accuracy of 99.67%even under moderate turbulence conditions,outperforming four other DL models.In addition,for experimental datasets collected from both atmospheric turbulence and underwater environments,the model consistently achieves recognition accuracies exceeding 90%,demonstrating strong generalization ability and a 77%reduction in parameter count compared to traditional convolutional neural network(CNN)-based DL models.We provide a new approach for deploying lightweight DL algorithms on resource-constrained embedded optical signal detection devices.
基金Supported by the National Natural Science Foundation of China(No:69872039)
文摘By defining fuzzy valued simple functions and giving L1(μ) approximations of fuzzy valued integrably bounded functions by such simple functions, the paper analyses by L1(μ)-norm the approximation capability of four-layer feedforward regular fuzzy neural networks to the fuzzy valued integrably bounded function F : Rn → FcO(R). That is, if the transfer functionσ: R→R is non-polynomial and integrable function on each finite interval, F may be innorm approximated by fuzzy valued functions defined as to anydegree of accuracy. Finally some real examples demonstrate the conclusions.
基金supported by the Key Research and Development Program of Jiangsu Province under Grant BE2022059-3,CTBC Bank through the Industry-Academia Cooperation Project,as well as by the Ministry of Science and Technology of Taiwan through Grants MOST-108-2218-E-002-055,MOST-109-2223-E-009-002-MY3,MOST-109-2218-E-009-025,and MOST431109-2218-E-002-015.
文摘Micro-expressions(ME)recognition is a complex task that requires advanced techniques to extract informative features fromfacial expressions.Numerous deep neural networks(DNNs)with convolutional structures have been proposed.However,unlike DNNs,shallow convolutional neural networks often outperform deeper models in mitigating overfitting,particularly with small datasets.Still,many of these methods rely on a single feature for recognition,resulting in an insufficient ability to extract highly effective features.To address this limitation,in this paper,an Improved Dual-stream Shallow Convolutional Neural Network based on an Extreme Gradient Boosting Algorithm(IDSSCNN-XgBoost)is introduced for ME Recognition.The proposed method utilizes a dual-stream architecture where motion vectors(temporal features)are extracted using Optical Flow TV-L1 and amplify subtle changes(spatial features)via EulerianVideoMagnification(EVM).These features are processed by IDSSCNN,with an attention mechanism applied to refine the extracted effective features.The outputs are then fused,concatenated,and classified using the XgBoost algorithm.This comprehensive approach significantly improves recognition accuracy by leveraging the strengths of both temporal and spatial information,supported by the robust classification power of XgBoost.The proposed method is evaluated on three publicly available ME databases named Chinese Academy of Sciences Micro-expression Database(CASMEII),Spontaneous Micro-Expression Database(SMICHS),and Spontaneous Actions and Micro-Movements(SAMM).Experimental results indicate that the proposed model can achieve outstanding results compared to recent models.The accuracy results are 79.01%,69.22%,and 68.99%on CASMEII,SMIC-HS,and SAMM,and the F1-score are 75.47%,68.91%,and 63.84%,respectively.The proposed method has the advantage of operational efficiency and less computational time.
基金the National Key R&D Program of China[grant numbers 2018YFB0504900 and 2018YFB0504904]the National Natural Science Foundation of China[grant numbers 42071325 and 42176183]+1 种基金LIESMARS Special Research Fundingthe“985 Project”of Wuhan University,and Special funds of State Key Laboratory for equipment.
文摘With a spatial resolution of 50 m,a revisit time of three days,and a swath of 950 km,the coastal zone imager(CZI)offers great potential in monitoring coastal zone dynamics.Accurate atmo-spheric correction(AC)is needed to exploit the potential of quantitative ocean color inversion.However,due to the band setting of CZI,the AC over coastal waters in the western Pacific region with complex optical properties cannot be realized easily.This research introduces a novel neural network(NN)AC algorithm for CZI data over coastal waters.Total 100,000 match-ups of HY-1 C CZI-observed reflectance at the top-of-atmosphere and Operational Land Imager(OLI)-retrieved high-quality remote sensing reflectance(Rrs)at the CZI bands are built to train the NN model.These reflectance data are obtained from the standard AC algorithm in the SeaDAS.Results indicate that the distributions of the CZI retrieved Rrs were consistent with the quasi-synchronous OLI data,but the spatial information from the CZI is more detailed.Then,the accuracy of the CZI data for AC is evaluated using the multi-source in-situ data.Results further show that the NN-AC can successfully retrieve Rrs for CZI and the coefficients of determination in the blue,green,red,and near-infrared bands were 0.70,0.77,0.76,and 0.67,respectively.The NN algorithm does not depend on shortwave-infrared bands and runs very fast once properly trained.
基金This project was supported by the National Natural Science Foundation of China (79970042).
文摘0-1 programming is a special case of the integer programming, which is commonly encountered in many optimization problems. Neural network and its general energy function are presented for 0-1 optimization problem. Then, the 0-1 optimization problems are solved by a neural network model with transient chaotic dynamics (TCNN). Numerical simulations of two typical 0-1 optimization problems show that TCNN can overcome HNN's main drawbacks that it suffers from the local minimum and can search for the global optimal solutions in to solveing 0-1 optimization problems.
基金Financial support of mechanical engineering center of excellence at Roudbar Azad University
文摘A neural network with a feed forward topology and Bayesian regularization training algorithm is used to predict the austenite formation temperatures （At1 and A13） by considering the percentage of alloying elements in chemical composition of steel. The data base used here involves a large variety of different steel types such as struc- tural steels, stainless steels, rail steels, spring steels, high temperature creep resisting steels and tool steels. Scatter diagrams and mean relative error （MRE） statistical criteria are used to compare the performance of developed neural network with the results of Andrew% empirical equations and a feed forward neural network with ＂gradient descent with momentum＂ training algorithm. The results showed that Bayesian regularization neural network has the best performance. Also, due to the satisfactory results of the developed neural network, it was used to investigate the effect of the chemical composition on Ac1 and At3 temperatures. Results are in accordance with materials science theories.
基金supported by the Natural Science Basic Research Plan in Shaanxi Province,China(No.2021JQ-214)the Fundamental Research Funds for the Central Universities,China(No.300102251101).
文摘The main task of this work is to design a control system for a small tail-sitter Unmanned Aerial Vehicle(UAV)during the transition process.Although reasonable control performance can be obtained through a well-tuned single PID or cascade PID control architecture under nominal conditions,large or fast time-varying disturbances and a wide range of changes in the equilibrium point bring nonlinear characteristics to the transition control during the transition process,which leads to control precision degradation.Meanwhile,the PID controller’s tuning method relies on engineering experiences to a certain extent and the controller parameters need to be retuned under different working conditions,which limits the rapid deployment and preliminary validation.Based on the above issues,a novel control architecture of L1 neural network adaptive control associated with PID control is proposed to improve the compensation ability during the transition process and guarantee the security transition.The L1 neural network adaptive control is revised to solve the multi-input and multi-output problem of the tail-sitter UAV system in this study.Finally,the transition characteristics of the time setting difference between the desired transition speed and the desired transition pitch angle are analyzed.
基金support provided by the China National Key Research and Development Program of China under Grant 2019YFB2004300the National Natural Science Foundation of China under Grant 51975065 and 51805051.
文摘Integrated with sensors,processors,and radio frequency(RF)communication modules,intelligent bearing could achieve the autonomous perception and autonomous decision-making,guarantying the safety and reliability during their use.However,because of the resource limitations of the end device,processors in the intelligent bearing are unable to carry the computational load of deep learning models like convolutional neural network(CNN),which involves a great amount of multiplicative operations.To minimize the computation cost of the conventional CNN,based on the idea of AdderNet,a 1-D adder neural network with a wide first-layer kernel(WAddNN)suitable for bearing fault diagnosis is proposed in this paper.The proposed method uses the l1-norm distance between filters and input features as the output response,thus making the whole network almost free of multiplicative operations.The whole model takes the original signal as the input,uses a wide kernel in the first adder layer to extract features and suppress the high frequency noise,and then uses two layers of small kernels for nonlinear mapping.Through experimental comparison with CNN models of the same structure,WAddNN is able to achieve a similar accuracy as CNN models with significantly reduced computational cost.The proposed model provides a new fault diagnosis method for intelligent bearings with limited resources.
文摘Due to the increasing variety of information and services carried by optical networks, the survivability of network becomes an important problem in current research. The fault location of OTN is of great significance for studying the survivability of optical networks. Firstly, a three-channel network model is established and analyzing common alarm data, the fault monitoring points and common fault points are carried out. The artificial neural network is introduced into the fault location field of OTN and it is used to judge whether the possible fault point exists or not. But one of the obvious limitations of general neural networks is that they receive a fixedsize vector as input and produce a fixed-size vector as the output. Not only that, these models is even fixed for mapping operations (for example, the number of layers in the model). The difference between the recurrent neural network and general neural networks is that it can operate on the sequence. In spite of the fact that the gradient disappears and the gradient explodes still exist in the neural network, the method of gradient shearing or weight regularization is adopted to solve this problem, and choose the LSTM (long-short term memory networks) to locate the fault. The output uses the concept of membership degree of fuzzy theory to express the possible fault point with the probability from 0 to 1. Priority is given to the treatment of fault points with high probability. The concept of F-Measure is also introduced, and the positioning effect is measured by using location time, MSE and F-Measure. The experiment shows that both LSTM and BP neural network can locate the fault of optical transport network well, but the overall effect of LSTM is better. The localization time of LSTM is shorter than that of BP neural network, and the F1-score of LSTM can reach 0.961566888396156 after 45 iterations, which meets the accuracy and real-time requirements of fault location. Therefore, it has good application prospect and practical value to introduce neural network into the fault location field of optical transport network.
基金supported by Jiangsu Social Science Foundation(No.20GLD008)Science,Technology Projects of Jiangsu Provincial Department of Communications(No.2020Y14)Joint Fund for Civil Aviation Research(No.U1933202)。
文摘In order to directly construct the mapping between multiple state parameters and remaining useful life(RUL),and reduce the interference of random error on prediction accuracy,a RUL prediction model of aeroengine based on principal component analysis(PCA)and one-dimensional convolution neural network(1D-CNN)is proposed in this paper.Firstly,multiple state parameters corresponding to massive cycles of aeroengine are collected and brought into PCA for dimensionality reduction,and principal components are extracted for further time series prediction.Secondly,the 1D-CNN model is constructed to directly study the mapping between principal components and RUL.Multiple convolution and pooling operations are applied for deep feature extraction,and the end-to-end RUL prediction of aeroengine can be realized.Experimental results show that the most effective principal component from the multiple state parameters can be obtained by PCA,and the long time series of multiple state parameters can be directly mapped to RUL by 1D-CNN,so as to improve the efficiency and accuracy of RUL prediction.Compared with other traditional models,the proposed method also has lower prediction error and better robustness.
文摘Neural stem cells(NSCs) and imprinted genes play an important role in brain development. On historical grounds, these two determinants have been largely studied independently of each other. Recent evidence suggests, however, that NSCs can reset select genomic imprints to prevent precocious depletion of the stem cell reservoir. Moreover, imprinted genes like the transcriptional regulator Zac1 can fine tune neuronal vs astroglial differentiation of NSCs. Zac1 binds in a sequence-specific manner to pro-neuronal and imprinted genes to confer transcriptional regulation and furthermore coregulates members of the p53-family in NSCs. At the genome scale, Zac1 is a central hub of an imprinted gene network comprising genes with animportant role for NSC quiescence, proliferation and differentiation. Overall, transcriptional, epigenomic, and genomic mechanisms seem to coordinate the functional relationships of NSCs and imprinted genes from development to maturation, and possibly aging.
文摘The hot deformation behavior of TI (18W-4Cr-1V) high-speed steel was investigated by means of continuous compression tests performed on Gleeble 1500 thermomechan- ical simulator in a wide range of tempemtures (950℃-1150℃) with strain rotes of 0.001s-1-10s-1 and true strains of 0-0. 7. The flow stress at the above hot defor- mation conditions is predicted by using BP artificial neural network. The architecture of network includes there are three input parameters:strain rate,temperature T and true strain , and just one output parameter, the flow stress ,2 hidden layers are adopted, the first hidden layer includes 9 neurons and second 10 negroes. It has been verified that BP artificial neural network with 3-9-10-1 architecture can predict flow stress of high-speed steel during hot deformation very well. Compared with the prediction method of flow stress by using Zaped-Holloman parumeter and hyperbolic sine stress function, the prediction method by using BP artificial neurul network has higher efficiency and accuracy.
文摘In developing Artificial Neural Networks(ANNs),the available dataset is split into three categories:training,validation and testing.However,an important problem arises:How to trust the predic-tion provided by a particular ANN?Due to the randomness related to the network itself(architecture,initialization and learning procedure),there is usually no best choice.Considering this issue,we provide a framework,which captures the randomness related to the network itself.The idea is to perform several training and test trials based on the Jackknife resampling method.Jackknife consists of iteratively deleting a single observation each time from the sample and recomputing the ANN on the rest of the sample data.Consequently,interval prediction is available instead of point prediction.The proposed method was applied and tested using pH,Ca and P data obtained by analyzing 118 georeferenced soil points.The results,based on the dataset size simulation,showed that 60%reduction in available dataset offers compatible accuracy in relation to full dataset,and therefore a higher cost of sampling in the field would not be necessary.The re-sampling method spatially characterizes the points of greater or lesser accuracy and uncertainty.The re-sampling method increased the success rate by using interval prediction instead of using the mean as the most probable value.Although we restrict it to the regression neural network model,the resampling method proposed can also be extended to other modern statistical tools,such as Kriging,Least Squares Collocation(LSC),Convolutional Neural Network(CNN),and so on.
基金supported by the National Basic Research and Development Program("973" Program),under contract No.2009CB421202the National Natural Science Foundation of China under contract No. 40706061the National High Technology Development Program ("863"Program),under contract Nos 2007AA12Z137 and 2008AA09Z104
文摘The geophysical model function （GMF） describes the relationship between backscattering and sea surface wind, so that wind vec- tors can be retrieved from backscattering measurement. The GMF plays an important role in ocean wind vector retrievals, its performance will directly influence the accuracy of the retrieved wind vector. Neural network （NN） approach is used to develop a unified GMF for C-band and Ku-band （NN-GMF）. Empirical GMF CMOIM and QSCAT-1 are used to generate the simulated training data-set, and Gaussian noise at a signal noise ratio of 30 dB is added to the data-set to simulate the noise in the backscat- tering measurement. The NN-GMF employs radio frequency as an additional parameter, so it can be applied for both C-band and Ku-band. Analyses show that the %predicted by the NN-GMF is comparable with the σpredicted by CMOIM and QSCAT-1. Also the wind vectors retrieved from the NN-GMF and empirical GMF CMOIM and QSCAT-1 are comparable, indicating that the NN-GMF is as effective as the empirical GMF, and has the advantages of the universal form.
文摘Audiovisual speech recognition is an emerging research topic.Lipreading is the recognition of what someone is saying using visual information,primarily lip movements.In this study,we created a custom dataset for Indian English linguistics and categorized it into three main categories:(1)audio recognition,(2)visual feature extraction,and(3)combined audio and visual recognition.Audio features were extracted using the mel-frequency cepstral coefficient,and classification was performed using a one-dimension convolutional neural network.Visual feature extraction uses Dlib and then classifies visual speech using a long short-term memory type of recurrent neural networks.Finally,integration was performed using a deep convolutional network.The audio speech of Indian English was successfully recognized with accuracies of 93.67%and 91.53%,respectively,using testing data from 200 epochs.The training accuracy for visual speech recognition using the Indian English dataset was 77.48%and the test accuracy was 76.19%using 60 epochs.After integration,the accuracies of audiovisual speech recognition using the Indian English dataset for training and testing were 94.67%and 91.75%,respectively.
基金Project supported by the National Key Research and Development Program of China(Grant No.2019YFB2205102)the National Natural Science Foundation of China(Grant Nos.61974164,62074166,61804181,62004219,62004220,and 62104256).
文摘Memristor-based neuromorphic computing shows great potential for high-speed and high-throughput signal processing applications,such as electroencephalogram(EEG)signal processing.Nonetheless,the size of one-transistor one-resistor(1T1R)memristor arrays is limited by the non-ideality of the devices,which prevents the hardware implementation of large and complex networks.In this work,we propose the depthwise separable convolution and bidirectional gate recurrent unit(DSC-BiGRU)network,a lightweight and highly robust hybrid neural network based on 1T1R arrays that enables efficient processing of EEG signals in the temporal,frequency and spatial domains by hybridizing DSC and BiGRU blocks.The network size is reduced and the network robustness is improved while ensuring the network classification accuracy.In the simulation,the measured non-idealities of the 1T1R array are brought into the network through statistical analysis.Compared with traditional convolutional networks,the network parameters are reduced by 95%and the network classification accuracy is improved by 21%at a 95%array yield rate and 5%tolerable error.This work demonstrates that lightweight and highly robust networks based on memristor arrays hold great promise for applications that rely on low consumption and high efficiency.
文摘With the rapid growth of the Internet in recent years, the ability to analyze and identify its users has become increasingly important. Authorship analysis provides a means to glean information about the author of a document originating from the internet or elsewhere, including but not limited to the author’s gender. There are well-known linguistic differences between the writing of men and women, and these differences can be effectively used to predict the gender of a document’s author. Capitalizing on these linguistic nuances, this study uses a set of stylometric features and a set of word count features to facilitate automatic gender discrimination on emails from the popular Enron email dataset. These features are used in conjunction with the Modified Balanced Winnow Neural Network proposed by Carvalho and Cohen, an improvement on the original Balanced Winnow created by Littlestone. Experiments with the Modified Balanced Winnow show that it is effectively able to discriminate gender using both stylometric and word count features, with the word count features providing superior results.
