Neural Network Research Research Articles

Rapid appearance quality of rice based on machine vision and convolutional neural network research on automatic detection system.

In the process of rice production and storage, there are many defects in the traditional detection methods of rice appearance quality, but using modern high-precision instruments to detect the appearance quality of rice has gradually developed into a new research trend at home and abroad with the development of agricultural artificial intelligence. In this study, we independently designed a fast automatic rice appearance quality detection system based on machine vision technology by introducing convolutional neural network and image processing technology. In this study, NIR and RGB images were generated into five-channel image data by superposition function, and image are preprocessed by combining the Watershed algorithm with the Otus adaptive threshold function. Different grains in the samples were labeled and put in the convolutional neural network for training. The rice grains were classified and the phenotype data were analyzed by selecting the optimal training model to realize the detection of rice appearance quality. The experimental results showed that the resolution of the system could reach 92.3%. In the detection process, the system designed with this method not only reduces the subjectivity problems caused by different detection environments, visual fatigue caused large sample size and the inspector's personal factors, but also significantly improves the detection time and accuracy, which further enhances the detection efficiency of rice appearance quality, and has positive significance for the development of the rice industry.

Klasifikasi Penyakit Daun Jagung Menggunakan Lightweight Convolutional Neural Network

Deep Neural Networks have achieved great success in many tasks and research especially in pattern recognition. However, high computing and large memory requirements prevent them from being applied to limited devices and require large costs. And Imbalance of data in Machine Learning refers to the unequal distribution of classes in a data set. This problem is often encountered in classification tasks where the distribution of classes within a given data set is not uniform. Given these problems, the researchers proposed the Light Neural Network method and added the Resampling method to Corn Leaf Disease data, to ease the computational process and balance the data. Researchers use several methods in machine learning which involve the use of knowledge that has been learned by existing models (Transfer Learning), including: MobileNet, EfficientNetB1 and Exception, as well as conducting comparisons between models. These models use the Depthwise Separable Convolution technique to reduce the number of parameters and convolution operations that need to be performed, thereby reducing the computational requirements and model complexity. The results of light neural network research and the combination of Resampling, the highest Transfer Learning model of 95% obtained from the EfficientNetB1 model than other Transfer learning models.

Research and application of wavelet neural network in electrical resistivity imaging inversion

In order to obtain high-quality electrical resistivity imaging (ERI) inversion results, this paper proposes a wavelet-based neural network inversion method. In addition, it is applied to 2D electrical resistivity tomography imaging. Firstly, a sample set suitable for wavelet neural network training is designed to optimize neural network parameters and improve the accuracy of electrical resistivity inversion. Secondly, a hybrid multi-layer wavelet neural network is designed, which uses Mexican-hat and Morlet wavelets as activation functions for different hidden layers, to improve the generalization ability and stability of the network. Finally, the superiority of the WNN is verified through two experiments and applied for inversion in field examples. The inversion results of the synthetic and field examples show that the introduced method is superior to other algorithms in terms of prediction accuracy and computational efficiency, which contribute to better inversion results.

ResNet and its application to medical image processing: Research progress and challenges

Background and objectiveDeep learning, a novel approach and subset of machine learning, has drawn a growing amount of attention from computer vision researchers in recent years. This method has drawn a lot of interest because of its extraordinary ability to interpret medical pictures, especially when combined with residual neural networks, which have helped to progress the field. MethodsIn this paper, the following research is carried out on the residual network. First, the research status of ResNet in the medical field is introduced. The fundamental idea behind the residual neural network is then explained, along with the residual unit, its many structures, and the network architecture. Second, four aspects of the widespread use of residual neural networks in medical image processing are discussed: lung tumor, diagnosis of skin diseases, diagnosis of breast diseases, and diagnosis of diseases of the brain. Finally, the main issues and ResNet's future development in the area of processing medical images are discussed. ResultsIn the area of medical graph processing, residual neural networks have made strides and have had success in the clinical auxiliary diagnosis of serious illnesses such as lung tumors, breast cancer, skin conditions, and cardiovascular and cerebrovascular diseases. ConclusionWe thoroughly sorted out the most recent developments in residual neural network research and their use in medical image processing, which serves as a crucial point of reference for this field of study. It offers a helpful reference for further promoting the application and research of the ResNet model in the field of medical image processing by summarising the application status and issues of the ResNet model in the field of medical image processing and putting forwards some future development directions.

Application of deep learning in smart agriculture research

With the breakthrough of deep neural network research, deep learning technology has developed rapidly and has been widely used in the area of smart agriculture and has achieved great accomplishment in target detection and recognition, numerical weather forecasting and data analysis. This paper summarizes the application of deep learning technology in three aspects: detection and identification of pests and diseases, weather forecast and breeding. At the same time, this paper further discusses the major challenges and future development directions of the application of deep learning technology in the area of smart agriculture, and provides a reference for further research into deep learning in the area of smart agriculture. Accurate identification of crop pests is an important prerequisite for pest prediction and control. Image recognition has recently emerged as the primary technological tool for pest management due to its benefits of high efficiency, cheap cost, and simple operation. Future trends in the realm of pest identification and detection include the fusion of transfer learning and deep learning. Deep learning has great advantages in computing efficiency, accuracy, portability, synergy, flexibility, and ease of use. It will be possible to change the traditional meteorological observation mode, accelerate and improve the processing of meteorological observation data. Future development trends include combining the benefits of deep learning and numerical forecasting. Deep learning has made many advances in the fields of genomics, transcriptomics, proteomics and synthetic biology, and can provide a powerful driving force for the fields of crop breeding and plant genomics. Standardized collection, processing, storage and management of data, and the establishment of an open and shared database are some of the future research directions.

Artificial intelligence to assist specialists in the detection of haematological diseases

Artificial intelligence, particularly the growth of neural network research and development, has become an invaluable tool for data analysis, offering unrivalled solutions for image generation, natural language processing, and personalised suggestions. In the meantime, biomedicine has been presented as one of the pressing challenges of the 21st century. The inversion of the age pyramid, the increase in longevity, and the negative environment due to pollution and bad habits of the population have led to a necessity of research in the methodologies that can help to mitigate and fight against these changes.The combination of both fields has already achieved remarkable results in drug discovery, cancer prediction or gene activation. However, challenges such as data labelling, architecture improvements, interpretability of the models and translational implementation of the proposals still remain. In haematology, conventional protocols follow a stepwise approach that includes several tests and doctor-patient interactions to make a diagnosis. This procedure results in significant costs and workload for hospitals.In this paper, we present an artificial intelligence model based on neural networks to support practitioners in the identification of different haematological diseases using only rutinary and inexpensive blood count tests. In particular, we present both binary and multiclass classification of haematological diseases using a specialised neural network architecture where data is studied and combined along it, taking into account the clinical knowledge of the problem, obtaining results up to 96% accuracy for the binary classification experiment. Furthermore, we compare this method against traditional machine learning algorithms such as gradient boosting decision trees and transformers for tabular data. The use of these machine learning techniques could reduce the cost and decision time and improve the quality of life for both specialists and patients while producing more precise diagnoses.

Hedonic preference of airline passengers

The purpose of this study is to determine which airlines Davao City-bound and outbound passengers prefer. Passenger information was acquired via survey questionnaires employing the Likert scale. In this study, non-experimental quantitative research employing a casual research method was employed. Using descriptive and informal research methodologies, this study investigated whether demographic variables and airline service quality influence the airline service choices of Davao City passengers. This investigation employs artificial neural network research to determine which factors influence airline preference. The majority of air carrier respondents are between the ages of 21 and 35, are predominantly female, and are employed by private companies. The airline preferences of passengers are influenced by price, cabin services, flight schedule, safety, on-time performance, and employee conduct. Important airline demographics included age, occupation, and sex. Price is the most significant criterion when selecting an airline, followed by flight schedule, on-time performance, and employee conduct. Flights should offer discounts to students and senior citizens. They may adjust fares based on the findings. To attract a broader age range of customers, airlines should strategically organize their marketing and focus on specific market niches. To increase sales, airlines may consider offering discounted, low-cost, or one-peso fares during certain times. To assist time-sensitive travelers in reaching their destinations in the event of airline delays or cancellations, alternative flights are advised. To gain a deeper understanding of airline carrier selection, research should be extended, employing a larger sample or alternative statistical methods.

Variable Binding for Sparse Distributed Representations: Theory and Applications.

Variable binding is a cornerstone of symbolic reasoning and cognition. But how binding can be implemented in connectionist models has puzzled neuroscientists, cognitive psychologists, and neural network researchers for many decades. One type of connectionist model that naturally includes a binding operation is vector symbolic architectures (VSAs). In contrast to other proposals for variable binding, the binding operation in VSAs is dimensionality-preserving, which enables representing complex hierarchical data structures, such as trees, while avoiding a combinatoric expansion of dimensionality. Classical VSAs encode symbols by dense randomized vectors, in which information is distributed throughout the entire neuron population. By contrast, in the brain, features are encoded more locally, by the activity of single neurons or small groups of neurons, often forming sparse vectors of neural activation. Following Laiho et al. (2015), we explore symbolic reasoning with a special case of sparse distributed representations. Using techniques from compressed sensing, we first show that variable binding in classical VSAs is mathematically equivalent to tensor product binding between sparse feature vectors, another well-known binding operation which increases dimensionality. This theoretical result motivates us to study two dimensionality-preserving binding methods that include a reduction of the tensor matrix into a single sparse vector. One binding method for general sparse vectors uses random projections, the other, block-local circular convolution, is defined for sparse vectors with block structure, sparse block-codes. Our experiments reveal that block-local circular convolution binding has ideal properties, whereas random projection based binding also works, but is lossy. We demonstrate in example applications that a VSA with block-local circular convolution and sparse block-codes reaches similar performance as classical VSAs. Finally, we discuss our results in the context of neuroscience and neural networks.

A Patch Information Supplement Transformer for Person Re-Identification

Extracting fine-grained features from person images has proven crucial in person re-identification (re-ID). Although the research of convolutional neural networks (CNN) has been very successful in person re-ID, due to the small receptive field and downsampling operation, the existing CNNs cannot solve the problem of information loss. The multi-head attention modules in transformer can solve the above problems well. However, since dicing operations destroy the spatial correlation between patches, transformer still loses some local features. in this paper, we propose the scheme of the patch information supplement transformer (PIT) to extract fine-grained features in the dicing stage. Patch pyramid network (PPN) is introduced to solve the problem of local information loss. This is accomplished by dividing the image into different scales through the dicing operation and adding them together from top to bottom according to the pyramid structure. In addition, we insert a learnable identity information-embedding module (IDE) to reduce the feature bias of clothing and camera perspective. Experiments verify the superiority and effectiveness of PIT compared to state-of-the-art methods.

Research of Convolutional Neural Network on Image Classification

With the progress of artificial intelligence, technology based on deep learning is becoming more and more mature, and the application of deep convolutional neural network for image classification has become a popular topic for researchers. The number of the structure of deep convolutional neural network for image classification is keep increasing, and its performance is consistently improving, gradually replace that of traditional methods. According to the process of model development and model optimization, this paper divides the convolutional neural network into two models: classical deep convolutional neural network model and attention mechanism deep convolutional neural network model. The construction methods and characteristics of various kinds of deep convolutional neural network models are comprehensively reviewed, and the performance of various classification models is compared and analyzed. Finally, the problems of deep convolutional neural network model are presented.

A Security-Enhanced and Privacy-Preserving Certificateless Aggregate Signcryption Scheme-Based Artificial Neural Network in Wireless Medical Sensor Network

With the recent advances in mobile sensing and beyond-5G technologies, wireless medical sensor networks (WMSNs) have become an emerging hot topic in smart healthcare. Under the condition of limited resources of medical sensor nodes (MSNs), ensuring the accuracy and integrity of a large number of patients’ medical data is a crux for security and data privacy issues. In order to overcome the limitations of medical sensors in communication and computing capabilities, and to achieve privacy in WMSN, we propose a low complexity certificateless aggregate signcryption (CL-ASC) scheme-based artificial neural network (NN) for safely achieving authentication and unforgeability of medical nodes. We use an adaptive sigma filter for denoising, adopt Levenshtein entropy for encoding, and design cryptographic protocol for signcrypting in the hidden layer of deep feedforward artificial NN (DFANN). The CL-ASC can authenticate all sensing bioinformation at once in a privacy-preserving way. Given the challenge of applying a discrete logarithm and computational Diffie-Hellman problems based on an elliptic curve, the proposed scheme satisfies the indistinguishability against chosen-ciphertext attacks (IND-CCA) and the adaptive chosen message attack (UF-CMA) security properties under the random oracle model (ROM). Finally, the benchmark evaluation shows that ours is ahead of similar state-of-the-art schemes, due to its low complexity and higher computational efficiency. Researchers interested in the provable security protocol design and framework of NNs can gain useful insights from this research and identify future cross domain of cybersecurity and NNs research directions based on the proposed scheme herein.

Research On Image Stylization Methods Based on Deep Network

For the purpose of promoting the research of deep neural networks, image stylization has received extensive attention. Researchers have proposed a huge number of image stylization methods based on deep neural networks. In this paper, image stylization methods are divided into two categories: image stylization method based on reference and based on domain. A series of stylization methods. Several classicals methods are summarized, and the performance of the two categories of stylization methods on the common data sets of image stylization tasks are compared respectively, the developments of image stylization in video, human face and brushwork are proposed, the current progress of image stylization and the future research direction are mentioned. At the end of the article, the full text is summarized and summarized.

Optimization of blasting parameters and prediction of vibration effects in open pit mines based on deep neural networks

Embedded systems in production equipment and Internet of Things (IoT) sensors on production lines are one of the elements that constitute an industrial cyber-physical system. In this paper, an in-depth study and analysis of the optimization of blasting parameters and prediction of vibration effects in open pit mines using deep neural network arithmetic are present. Based on the deep neural network research and analysis of the relationship between blasting parameters and rock fragmentation, a prediction model for blasting parameters and fragmentation for the East Open Pit Mine was established, and sensitivity analysis was performed on blasting parameters, and the unit consumption of explosives and the perforation rate were established. It was found that the average relative errors of both numerical simulation results and depth prediction results were no more than 10%, while the average relative errors of Sadowski's formula prediction results were more than 20%. The results show that the neural network optimized by a genetic algorithm and the numerical simulation has the highest accuracy in predicting the blasting result parameters. The research model and results obtained in this paper can be used as a reference guide for engineering practice.

SPAYK: An environment for spiking neural network simulation

In research areas such as mobile robotics and computer vision, energy and computational efficiency have become critical. This has greatly increased interest in high-efficiency neuromorphic hardware and spiking neural networks. Because neuromorphic hardware is not yet widely available, spiking neural network studies are conducted by simulations. There are numerous simulators available today, each designed for a specific purpose. In this paper, a novel and open source package (SPAYK) for simulating spiking neural networks is presented. SPAYK has been proposed to speed up spiking neural network research. In the majority of simulators, networks are expressed with differential equations and require advanced neuroscience knowledge since such simulators are generally designed for brain and neuroscience research. SPAYK, on the other hand, is specifically designed as a framework to easily design spiking neural networks for practical problems. SPAYK is an easy-to-use Python package. There are three fundamental classes in the core: the model class for creating neuron groups, the organization class for simulating tissues, and the learning class for synaptic plasticity. While developing and testing the SPAYK environment, various experiments were carried out. This study includes three of these experiments. In the first experiment, we investigated the behavior of a group of Izhikevich neurons for visual stimuli. Also, a single Izhikevich neuron has been trained to respond to a particular label in a supervised manner with synaptic plasticity. In the second experiment, a well-known experiment was repeated to validate SPAYK. In this experiment, a neuron trained by synaptic plasticity can recognize repetitive patterns in a spike train. In the third experiment, a similar neuron was simulated with stimuli with multiple labels adapted from the MNIST dataset. It has been shown that the neuron can classify a particular label by synaptic plasticity. All these experiments and the SPAYK environment are presented as open-source tools.

Development and research of triangle-filter convolution neural network for fuel reloading optimization of block-type HTGRs

The problem of fuel reloading optimization is very demanding, which requires to search for the optimal suitable core configuration within a very huge solution space. To solve this problem, updating of optimization solutions and quantitative evaluation of core configurations are two main steps. With the development of computing science, the meta-heuristic algorithms (MHAs) turn out to be effective for updating optimization solutions; while the artificial neural networks (ANNs) proves to be practical for the other. However, there is few research for the application of convolution neural networks (CNNs). For any given image-like inputs, CNNs could quickly calculate the outputs mainly by convolution operators, using rectangle filters to scan the full inputs. However, the geometry of some reactors is not rectangle so that using rectangle filters fails to traverse. Aiming at solving the problems of fuel reloading optimization for block-type HTGRs, the triangle-filter convolution neural networks (TFCNNs) are developed. At the same time, uniform method and ring-ranked method are also developed for determining the weights of TFCNNs. Combined with genetic algorithm and particle swarm optimization, two wide-applied MHAs, TFCNNs are applied to solve the problems of fuel reloading optimization for block-type HTGRs. The results show that the TFCNNs could greatly predict the optimization parameters of testing data set with accuracy higher than 90%. In addition, comparing to DRAGON V4 program, the calculation time of TFCNNs is even less than 1.30% of it, while comparing to BP-ANNs, the ratio of feasible solutions by TFCNNs is evidently higher.

Brain-Inspired Spiking Neural Network Using Superconducting Devices

Based on recent research in artificial neural networks, researchers have focused on topics from brain-like computing based on the Von Neumann architecture to brain-inspired computing based on the integration of storage and calculation due to the large energy dissipation. Inspired by biological neural networks, people use superconducting devices, which have nonlinear dynamic characteristics, to construct synapse and neuron circuits. However, it is challenging to build large-scale superconducting spiking neural networks due to fan-ins and fan-outs of superconducting devices. In this paper, we present a criterion to scale up the superconducting neural network. Then, we present a three-layer fully connected superconducting spiking neural network construction scheme. Inspired by the biological neural network, we present a new training method based on frequency coding. After only 100 training iterations with one grayscale image of a digit, the learned three-layer fully connected superconducting spiking neural network has a recognition accuracy rate of 86.1% on the digit class. Moreover, the power dissipation of one spiking event in a superconducting synapse is <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$23.2{\boldsymbol{zJ}}$</tex-math></inline-formula> , which shows that the network has outstandingly high efficiency over existing biologically inspired neural networks.

Memory association circuit based on memristor for neural networks

Memristors have been favored in artificial intelligence, neural networks, and chaotic circuits, especially in neural synapses because of their unique advantages such as resistance variability, nonvolatile nature, and nanometer size. Benefits such as integration scale and low power consumption contribute toward simulating the biological synaptic function. Compared with memory association circuits using traditional CMOS transistors, memristors will reduce the complexity of the circuit and the power consumption. Therefore, it is greatly promising to use memristors as synapses to construct neural networks to mimic human brain functions. This paper successfully establishes a recognition circuit based on memristors to recognize some characteristics (size, color, shape, and smooth) of fruits, which is a learning function. After a few seconds, the output signal voltage drops, and this is a forgetting function. Through the establishment of a recognition circuit, the neural network and human complex behavior were simulated. This work lays the foundation for further research of human neural networks.

Dynamic Behavior Analysis and Synchronization of Memristor-Coupled Heterogeneous Discrete Neural Networks

Continuous memristors have been widely studied in recent years; however, there are few studies on discrete memristors in the field of neural networks. In this paper, a four-stable locally active discrete memristor (LADM) is proposed as a synapse, which is used to connect a two-dimensional Chialvo neuron and a three-dimensional KTZ neuron, and construct a simple heterogeneous discrete neural network (HDNN). Through a bifurcation diagram and Lyapunov exponents diagram, the period and chaotic regions of the discrete neural network model are shown. Through numerical analysis, it was found that the chaotic region and periodic region of the neural network based on DLAM are significantly improved. In addition, coexisting chaos and chaos attractors, coexisting periodic and chaotic attractors, and coexisting periodic and periodic attractors will appear when the initial value of the LADM is changed. Coupled by a LADM synapse, two heterogeneous discrete neurons are gradually synchronized by changing the coupling strength. This paper lays a good foundation for the future analysis of LADMs and the related research of discrete neural networks coupled by LADMs.

Photonic multiplexing techniques for neuromorphic computing

Abstract The simultaneous advances in artificial neural networks and photonic integration technologies have spurred extensive research in optical computing and optical neural networks (ONNs). The potential to simultaneously exploit multiple physical dimensions of time, wavelength and space give ONNs the ability to achieve computing operations with high parallelism and large-data throughput. Different photonic multiplexing techniques based on these multiple degrees of freedom have enabled ONNs with large-scale interconnectivity and linear computing functions. Here, we review the recent advances of ONNs based on different approaches to photonic multiplexing, and present our outlook on key technologies needed to further advance these photonic multiplexing/hybrid-multiplexing techniques of ONNs.

All-optical reservoir computing system based on polarization dynamics

Reservoir computing (RC) is a simplified recurrent neural network and can be implemented by using a nonlinear system with delay feedback, thus it is called delay-based RC. Various nonlinear nodes and feedback loop structures have been proposed. Most of existing researches are based on the dynamical responses in intensity of the nonlinear systems. There are also a photoelectric RC system based on wavelength dynamics and an all-optical RC based on the phase dynamics of a semiconductor laser with optical feedback, as well as so-called polarization dynamics of a vertical cavity surface emitting laser (VCSEL). However, these VCSEL-RCs actually are based on the intensity dynamics of two mutually orthogonal polarization modes, or polarization-resolved intensity dynamics. The RC based on rich dynamical responses in polarization has not yet been found. A semiconductor optical amplifier (SOA) fiber ring laser can produce rich dynamical states in polarization, and is used in optical chaotic secure communication and distributed optical fiber sensing. To further expand the application of polarization dynamics of the SOA fiber ring laser and open up a new direction for the research of optical RC neural network, an all-optical RC system based on polarization dynamics of the ring laser is proposed. The ring laser is used as the reservoir, and the SOA as the nonlinear node. After the input signal is masked according to a synchronization scheme, it is injected into the reservoir by intensity modulation for a continuous wave generated by a superluminescent light emitting diode (SLED). The dynamical response in polarization of the ring laser is detected by a polarizer and a photodetector. The influences of the SOA operation current, output power of the SLED and attenuation of a variable optical attenuator (VOA) in the fiber loop on the polarization dynamic characteristic (mainly referring to the output degree of polarization) of the laser are analyzed experimentally. The fading memory and nonlinear response of the RC system based on the polarization dynamic response and intensity dynamic response are compared experimentally. The influences of output power of the SLED and attenuation of the VOA on fading memory, consistency and separation of the RC system based on the two kinds of dynamic responses are investigated experimentally. Thus the range of the VOA attenuation is determined. The network performance of the polarization dynamics RC system is evaluated by processing a Santa Fe time series prediction task and a multi-waveform recognition task. The normalized mean square error can be as low as 0.0058 for the time series prediction task, and the identification rate can be as high as 100% for the recognition task under the appropriate system parameters and only 30 virtual nodes. The experimental results show that the polarization dynamics RC system has good prediction performance and classification capability, which are comparable to the existing RC system based on intensity dynamics of the ring laser. The system can be expected to process two tasks in parallel when the polarization dynamics and intensity dynamics are used at the same time.