Search

In the field of abstract text summarization, architectures based on encoder-decoder frameworks are widely applied to sequence-to-sequence generation tasks and can effectively handle sequences of unlimited length. Subsequently, the transformer model use a global attention mechanism, allowing encodings at different distances to mutually interact, greatly enhancing the model’s contextual awareness. However, this context-awareness is global, requiring the model to additionally learn to extract different levels of information to increase understanding. We improve the structure of the model to introduce prior knowledge so that it can learn from the global and local information and enhance the model’s understanding ability. This paper proposes global information-aware encoding and local information-aware encoding, which enhance the understanding of documents from coarse-grained and fine-grained perspectives respectively. Global encoding adds an extra feature to the encoder stage and performs attention with the document, generating a global summary encoding of the entire document to guide the generation of the summary content. Local encoding is to perform local convolution on the features extracted by the encoder, use prior knowledge to extract local features of the document and enable the model to quickly extract local detail information. Experiments show that the improved model proposed in this paper has higher rouge scores than the baseline model on the LCSTS and CSL datasets, and also has advantages over some mainstream models. The generated summaries are more accurate and informative. The code is available on github. url: https://github.com/keptupp/A-text-summarization-approach-to-enhance-global-and-local-information-awareness-of-transformer.

The color and texture of camouflaged objects closely resemble their backgrounds, making it difficult to distinguish between the target and the background. Additionally, camouflaged environments are complex, with some camouflaged objects overlapping significantly with the background, further complicating the identification of camouflaged target areas. To address these challenges, we present a novel double-branch camouflaged object detection algorithm, termed TDNet, inspired by biological vision and based on the classical convolutional neural network ResNet50. TDNet employs a two-stage target detection structure that utilizes location segmentation to accurately identify detection targets. Specifically, we introduce an internal feature interaction mechanism (IFE) that enhances camouflaged target feature representations through multi-dimensional spatial feature enhancement and multi-channel feature interaction, enabling the rough segmentation of camouflaged regions. To mitigate the issue of blurred boundaries of camouflaged targets, we devise a cross-layer feature fusion mechanism (CFU) to refine the borders of camouflaged regions. Moreover, we employ a multi-supervised approach to constrain the model’s degradation process, ensuring the accuracy of camouflaged object detection. Experimental results demonstrate that our proposed method outperforms nine classical camouflaged object detection algorithms, achieving state-of-the-art performance. We will release the code and the test maps at https://github.com/zc199823/TDNet.

In recent years, jute, as an important natural fiber crop, has become more and more significant in the production process of insect pests, causing serious harm to agricultural production. Especially in the field of crop pest identification with complex backgrounds, fuzzy features, and multiple small targets, the lack of datasets specifically for jute pests has led to the large limitations of traditional pest identification models in terms of generalization. At the same time, the research on models specifically for jute pest detection is still in its infancy. To solve this problem, we constructed a large-scale image dataset containing nine types of jute pests, which was highly targeted and could effectively support model training and evaluation. In this study, we developed a deep convolutional neural network model based on YOLOv7, namely JutePest-YOLO. The model has optimized the Backbone, Head, and loss functions of the baseline model, and introduced the new ELAN-P module and P6 detection layer, which effectively improved the model’s ability to identify jute pests in complex backgrounds. The experimental results showed that compared with the baseline model, the Precision, Recall, and F1 scores of the JutePest-YOLO model were improved by 3.45%, 1.76%, and 2.58%, respectively; the mAP@0.5 and mAP@0.5:0.95 was improved by 2.24% and 3.25%, and the overall model’s computation (GFLOPS) was reduced by 16.05%. Compared to other advanced methods such as YOLOv8s, JutePest-YOLO has achieved superior performance in terms of detection accuracy, with a precision of 98.7% and mAP@0.5 reaching 95.68%. As a result, JutePest-YOLO not only achieved significant improvement in recognition accuracy but also optimized computational efficiency. It’s a high-performance, lightweight solution for jute pest detection.

As a nationally protected land resource, farmland plays a crucial role in agriculture production and food safety, making the quality of soil and environmental health critically important. Therefore, studying the extent of soil heavy metal pollution in farmland is of great significance for understanding the growth environment of food crops and protecting agricultural land resources. This study addresses the challenge of accurately, quickly, and conveniently assessing the extent of soil heavy metal pollution across an entire research area using a limited number of soil samples. To tackle this issue, a novel soil heavy metal pollution risk hybrid intelligent evaluation model (HIEM) is proposed. The HIEM utilizes the Semi-Supervised Bayesian Regression (Semi-BR) model, trained through Bayesian Co-training, to predict the soil heavy metal content at unsampled points. It employs an improved Multiple Kernel Support Vector Machine (MKSVM) model to evaluate the pollution status of the soil. Additionally, Geographic Information System (GIS) techniques are employed for spatial analysis of the pollution situation in the research area. The study focuses on eight soil heavy metals: As, Cd, Cr, Hg, Pb, Zn, Cu, and Ni. The experimental verification of the model was conducted using field sampling data from the major agricultural areas of Huangpi and Xinzhou in Wuhan, Hubei Province, China. The experimental results show that the eastern region of Huangpi District is more severely contaminated, particularly the central area in the northeast, with moderate to high pollution levels. The hybrid intelligent evaluation model achieves an average accuracy of 96.66% in assessing single-factor pollution of the eight soil heavy metals and an overall evaluation accuracy of 97.42%. The hybrid intelligent evaluation model is able to accurately fit traditional single-factor index methods and Nemerow comprehensive pollution index method. The Geographic Information System representation reveals a consistent distribution trend of soil heavy metal pollution reflected by the hybrid intelligent evaluation model with the results obtained from single-factor index and Nemerow comprehensive pollution index evaluation, indicating the feasibility of using this evaluation method for assessing the risk of soil heavy metal pollution. The conclusion shows that the hybrid intelligent evaluation model needs at least 639 sets of sample data to achieve the highest accuracy when assessing the risk of soil heavy metal contamination in an area of about $3.7\times 10^{4}\,\,hm^{2}$ , and this paper provides a reference to solve the problem of realizing high-precision risk assessment of heavy metal contamination of agricultural soils in the case of small samples. This study is of great practical significance for soil pollution investigation, soil quality assessment and other practical work.

Music is composed of a set of regular sound waves, which are usually ordered and have a large number of repetitive structures. Important notes, chords, and music fragments often appear repeatedly. Such repeated fragments (referred to as motifs) are usually the soul of a song. However, most music generated by existing music generation methods can not have distinct motifs like real music. This study proposes a novel multi- encoders model called Motif Transformer to generate music containing more motifs. The model is constructed using an encoder-decoder framework that includes an original encoder, a bidirectional long short term memory-attention encoder (abbreviated as bilstm-attention encoder), and a gated decoder. Where the original encoder is taken from the transformer’s encoder and the bilstm-attention encoder is constructed from the bidirectional long short-term memory network (BILSTM) and the attention mechanism; Both the original encoder and the bilstm-attention encoder encode the motifs and input the encoded information representations to the gated decoder; The gated decoder decodes the entire input of the music and the information passed by the encoders and enhances the model’s ability to capture motifs of the music in a gated manner to generate music with significantly repeated fragments. In addition, in order to better measure the model’s ability of generating motifs, this study proposes an evaluation metric called used motifs. Experiments on multiple music field metrics show that the model proposed in this study can generate smoother and more beautiful music, and the generated music contains more motifs.

Cotton harvest can be increased by having real-time information on the state of cotton aphid populations. However, traditional cotton aphid monitoring relies on ground sample methods supported by models such as linear regression, resulting in low forecast accuracy. Therefore, this paper purposes to enhance the precision of the remote sensing prediction model by investigating the cotton aphid prediction model construction approach. We explored the effectiveness of the XGBoost algorithm combined with the GWO algorithm and SVR method for cotton aphid prediction relying on vegetation indices derived from UAV multispectral photography. Originally, 12 indices related to cotton aphids were calculated by UAV multispectral reflectance. Additionally, the optimal index combination for pest prediction was determined utilizing analysis of correction and two-way ANOVA, combined with the XGBoost algorithm. Furthermore, a pest prevalence prediction model for cotton aphids was constructed via the SVR methodology associated with the optimal catalog combination, and the model was optimized using the GWO algorithm. Compared with the seven algorithms, experimental results demonstrate that the MSE and MAE of the XGBoost-GWO-SVR model are reduced by 90.20% and 70.36% (SVR), 90.14% and 70.26% (XGBoost-SVR), 7.47% and 0.14% (XGBoost-GA-SVR), 5.80% and 0.11% (XGBoost-PSO-SVR), 12.06% and 58.95% (LR), and 84.77% and 89.22% (BPNN), whereas the $R^{2}$ is increased by 22.5% (SVR and XGBoost-SVR), 0.3% (LR), and 12.51% (BPNN). The $R^{2}$ of the prediction model of XGBoost-SVR combined with GWO, PSO, and GA is not significantly different. Among these models, the XGBoost-GWO-SVR obtained the highest $R^{2}$ of 0.980 and the lowest MAE of 2.838.

Urban sound event detection can automatically preload relevant information for a robot to ensure that it can be applied to various scene-activity tasks. To address the limitations of timbre similarity and scene recognition by audio collection devices, a fusion model based on the self-attention mechanism is proposed in this paper. The model consists of scattering transform and self-attention model. The scattering transform computes modulation spectrum coefficients of multiple orders through cascades of wavelet convolutions and modulus operators. It is learnable compared with Mel-scale Frequency Cepstral Coefficients (MFCC), and can be used to better restore the semantic features of some sound scenes with similar timbres. The transformer has an outstanding effect on Natural Language Processing (NLP) owing to its self-attention mechanism. In this paper, the self-attention mechanism in its encoder was used in the model, mainly to make the feature granularity consistent to refine the features. In addition, Focal Loss function was adopted in the model to curb the sample distribution imbalance. The Google Command and ESC-50 were used to supplement the scene categories of dataset UrbanSound8K. The model parameters of the learnable filters that performed well on the dataset UrbanSound8K were preserved to fine-tune the other two datasets with insufficient data volume and more target categories. The length of slice duration was further explored the in the model. The experimental results show that the model can achieve better performance in a large range of scene models.

To improve the accuracy of soil heavy metal content prediction, this study proposes a dynamic neural network optimization model (DNNOM). The model is based on a radial basis function neural network (RBFNN). The weights and bias of the output layer of the RBFNN were generated using an adaptive dynamic genetic optimization algorithm (ADGOA), and the center point of the hidden layer of the RBFNN was determined using an efficient density peak clustering algorithm (EDPC). An adaptive variance measure (AVM) was then used to generate the width vector of RBFNN hidden layer. The model was applied to the predict soil heavy metal content in six new urban areas in Wuhan. Through comparison with support vector machine(SVM), light gradient boosting machine(LightGBM), RBFNN, and genetic algorithm optimizes the radial basis function neural network(GA-RBFNN), the experimental results demonstrate that the DNNOM is closer to the real value than the other four models, and the four error indicator values are also significantly lower than those of the other comparison models, which have higher prediction accuracy. Especially when compared with RBFNN, the MAPE and SMAPE of DNNOM decreased by 3.98% and 3.9%, respectively.

The micro-motion target echoes can be regarded as the accumulation of a few strong scattering points echoes, which are naturally sparse. Therefore, the compressed sensing (CS) reconstruction method can be used to analyze the incomplete micro-Doppler signal and extract the micro-motion features. Traditional CS reconstruction algorithms are time-consuming and sensitive to the selection of model parameters, limiting the performance of micro-motion feature extraction. This paper proposes a deep unfolded method to achieve the reconstruction of the time-frequency representation (TFR) of incomplete micro-Doppler signals. First, the joint time-frequency (JTF) transform is used to construct the CS model and the Training data is obtained according to the model. Then, the alternating direction method of multipliers (ADMM) algorithm is constructed into an iterative network named ADMM-net corresponding to the iterative solution process. An auxiliary loss is designed into the ADMM-net called AL-ADMM-net to improve the quality of reconstruction. The AL-ADMM-net is trained by the training data to learn the optimal parameters. Furthermore, an AL-ADMM-net iterative method is proposed when the micro-Doppler signal has phase corruption. Simulations are given to prove the effectiveness of the proposed method.

To address the problem that existing deep learning methods are not sufficiently accurate to detect rice pests with changeable shapes or similar appearances, a self-attention feature fusion model for rice pest detection (SAFFPest) was proposed. The model was based on VarifocalNet. First, a deformable convolution module was added to the feature extraction network, to improve the feature extraction ability of pests with changeable shapes. Second, by obtaining the balance features of multiple feature maps, the self-attention mechanism was introduced to refine the balance feature, in order to better restore the semantic information of some pests with similar appearances. Subsequently, the group normalization method was used to replace the batch normalization method in the original model, to reduce the impact of batch size on model training. The IP102 rice pest dataset was used to train and verify this model. The experimental results showed that the model can accurately detect nine kinds of rice pests, such as rice leaf rollers and rice leaf caterpillars. Compared with FasterRCNN, RetinaNet, CP-FCOS, VFNet and BiFA-YOLO, the mean average precision of the model improved by 33.7%, 6.5%, 4.5%, 2.9% and 2% respectively.

Vortex electromagnetic (EM) wave has shown great potential in the field of radar imaging. Under the assumption that the target is located at the grid point, the compressed sensing (CS) algorithm is applied to vortex EM imaging and achieves high imaging quality. However, the off-grid target in the actual scene deviates from the grid points, resulting in the basis mismatch problem. This problem will lead to the mismatch between the assumed measurement matrix and the real echo signal and the degradation of the CS imaging performance. In order to solve this problem, a novel two-dimensional (2-D) joint off-grid imaging method is proposed by considering the gridding error as a multiplicative matrix and separating it via series expansion. This method compensates for errors in 2-D matrix directly instead of converting the 2-D matrix into the 1-D vector, which can not only solve the off-grid problem, but also reduce the storage of data acquisition and processing. Furthermore, the 2-D iterative adaptive algorithm (2DIAA) for the 2-D model is adopted to achieve better reconstruction performance and lower computational complexity. Finally, Simulation results show the effectiveness and robustness of the algorithm.

In outdoor sports such as alpine skiing, athletes hope to obtain information such as movement trajectory and speed to achieve scientific training. The portable sensor has a sense of restraint and affects the athlete’s performance. Cameras have difficulty capturing the whole race in an outdoor environment. For alpine skiing needs, we designed an alpine skiing information acquisition system, using unmanned aerial vehicle (UAV) and ground cameras to obtain information on the whole path of the race. The environment of the alpine skiing field, the high-speed curvilinear movement of athletes and the flight characteristics of UAVs cause challenges in the detection and tracking of athletes by UAVs. To realize the detection and tracking of alpine skiing by UAVs, we propose a detection algorithm by combining neural networks and correlation filters. A tracking confidence score based on the motion distance of the interval frame is defined. When the tracking confidence of the neural network detector is less than the threshold, the correlation filter is used to expand the recognition range and realize redetection. In addition, we created and annotated a fresh alpine skiing dataset. We compare our algorithm with four advanced algorithms on the UAV123 dataset. Three methods are used to evaluate the performance of our method on the alpine skiing sequence dataset. From the simulation results, our algorithm outperforms the comparison methods in terms of accuracy and robustness. Therefore, our algorithm has application value in the scientific training of alpine skiers.

In the application of regression prediction through big data technology, the error between the predicted value and the true value is often large. In order to reduce the error of data prediction, this paper proposes an Intelligent Data Prediction (IDP) scheme for Smart Service. It uses Least Squares Support Vector Machine (LSSVM) as the basic prediction model. Since there is no standard procedure for determining the main parameters of LSSVM, an improved Particle Swarm Optimization (MBPSO) algorithm is used to simultaneously optimize the parameters of LSSVM. The main disadvantage of PSO is precocity due to the disappearance of population diversity. Based on this, Improvement strategy of MBPSO aims to continuously generate “More” and “Better” particles. First, in order to avoid the early disappearance of particle diversity, MBPSO re-adjusted the inertia weight and learning factor. Secondly, a renewable access strategy is proposed to allow a part of the disappeared population to regenerate. Finally, the method of global optimal adjustment is introduced to help particles find the optimal flight direction. In order to verify the effectiveness of MBPSO, 9 test functions are used to test the algorithm performance. The results show that MBPSO's optimization speed, best and mean all perform best. Taking the farmland soil heavy metal data sets of Dongxihu District and Hannan District of Wuhan City as examples of application, the content of heavy metals Cr and Pb in the soil was predicted. The results show that the predicted value of IDP is closer to the actual value, and the three error index values are significantly lower than other models. Especially in the prediction of Pb content, compared with the LSSVM model, the prediction errors of the two regions are reduced by 25.67% and 20.70% respectively. We can conclude that the proposed IDP scheme has practical significance in data prediction.

Identification of the seam to be welded in automatic welding is an important challenge, especially on a complex 3D surface. At present, most studies on weld extraction are for butt joints and fillet joints. Few studies focus on D-type welds, which connect to pipelines or pressure vessels. In this paper, a method of identification and extraction of pipeline joint welds based on 3D point clouds is proposed. First, the problem that the surface of pipe fittings cannot be effectively stitched owing to lack of markings is solved by putting marker points on the surface to allow accurate splicing of the point cloud. Second, according to the contour feature information of the workpiece, a processing method for D-type welding seams is proposed. Finally, a boundary extraction algorithm based on the point cloud normal is used to obtain the point cloud of the weld. The experimental results show that this method can effectively extract the weld. The running time for the point cloud weld extraction method is less than half that of the existing methods. The accuracy of the proposed method is improved in all three directions: by 28.2% in the X direction, 19.1% in the Y direction and 16.0% in the Z direction. Moreover, this method is convenient for transforming weld coordinates to robot-based coordinates for the automatic welding of spatially complex welds.

Recently, edge caching has emerged as a new data sponsoring scheme, where the devices on the edge network, e.g., 4G/5G base stations, cache video contents and directly delivery them to mobile video users. Such caching schemes have significantly decreased the backhaul congestion, reduce the delivery latency, thus attract more mobile video users and monetize a huge number of videos. In this paper, we consider a hybrid caching schemes system, which combines edge cache sponsoring (ECS) and traditional cellular data sponsoring (CDS), where the content providers display advertisements to mitigate the cellular data downloading expense and provide attractive contents for mobile video users. In the system, we investigate the cooperative scenario and the competitive scenario between the two sponsoring schemes, and a mobile video user can select one scheme or neither of them for the requests of a video. In the first scenario, we focus on achieving the maximal total benefit of content providers by optimizing CDS and ECS schemes together. In the second scenario, we separately optimize ECS and CDS schemes and maximize the benefits of corresponding content providers. To illustrate the effectiveness of the two scenarios, we formulate the system as a two-stage game: 1) in the first stage, content providers (i.e., leaders) choose the sponsor effort in sponsor schemes (cooperatively or competitively); 2) in the second stage, mobile video users (i.e., followers) select their sponsor preferences. We conduct numerical results to explore the sub-game perfect equilibrium and find that the joint utilization of the two sponsor schemes can benefit the mobile video users, i.e., when ECS and CDS compete with each other, MUs can benefit 36% ~ 140% more than the case where there exists only one sponsor scheme. While when CPs cooperate with each other, their total payoff is maximized, the mobile video users' payoff also increases. Moreover, we find that ECS can benefit the content providers more than CDS when sponsor revenue is high, and this indicates that ECS is a promising sponsor scheme for the high-value contents.

Parameter estimation of variogram models is an important problem in geostatistics and environmental engineering. Most of existing works aim to estimate parameters of variogram single models, while neglecting the parameter estimation of variogram nested model. Most recently, the evolutionary algorithms(EA), including genetic algorithm(GA), are exploited to calculate the parameters of variogram model, which can obtain a more accurate solution. These methods have some hyper-parameters to set and suffer from the well-recognized premature convergence and slow global convergence problem of EA. In this paper, a double elite co-evolutionary genetic algorithm(DECGA) and deep reinforcement learning(dueling DQN) was introduced to estimate the parameters of variogram single or nested models so as to achieve better generalization performance. The DECGA can get the global optimal solution faster than GA with the help of dueling DQN, which can set the hyper-parameters according to the state of DECGA. To verify the effectiveness of the proposed method(DDQNGA), we conduct experiments on the agricultural heavy metal database. Experimental results demonstrate that our method can obtain parameter estimation more accurately. The method proposed in this paper have a certain practical value in the field of geostatistics and environmental engineering.

As a crucial task in conversation systems, response generation for multi-turn conversation aims to generate a coherent, informative and diverse response according to the conversation context. Existing models of this task are limited in their ability to capture long-term dependencies within and between utterances and to identify pertinent or important information in the context. Inspired by the Transformer neural network based solely on attention mechanisms recently proposed in machine translation, we propose a novel hierarchical structured multi-head attention network (HMAN) model to address both problems. Specifically, the context sequences are encoded in a hierarchical structure in which multi-head self-attention is first employed to compute the representation matrix of each utterance, and then, these utterance matrices are integrated to form a context representation with the complicated dependencies learned using multi-head attention. The experimental results on two public conversation datasets demonstrate that our proposed model significantly outperforms several state-of-the-art baselines with respect to both automatic evaluation and human evaluation and can generate more diverse and informative responses.

The prediction of soil heavy metal content is an important part of the management of soil heavy metal pollution, but it is often ignored. At present, there are few studies on the prediction of soil heavy metal content, and it is an urgent problem to choose an efficient method for soil heavy metal content prediction. In this paper, a collaborative compound neural network model (CCNN) was put forward to predict the soil heavy metal content, this model uses wavelet neural network (WNN) as the basic prediction model, and at the same time proposes a parallel bird swarm algorithm (PBSA) to solve the parameter optimization problem of WNN, based on the bird swarm algorithm (BSA), the PBSA not only increases the gathering behavior of individual, but also adopts sine transformation based on fitness difference ratio to carry out the following behavior of beggars to improve the global optimization ability, besides that, the acceptance criterion is used to compare the fitness of individuals after updating to avoid falling into a local optimum. Soil heavy metal content data from Yinchuan city of Ningxia and six new urban areas in Wuhan, China are used to make prediction experiments respectively, through compare with support vector machine (SVM), radial basis function neural network (RBFNN), WNN and bird swarm algorithm optimizes wavelet neural network (BSA-WNN), the experimental results demonstrate that the predicted value of the CCNN is closer to the actual value and has better prediction performance.

Humans and other animals can easily perform self-localization by means of vision. However, that remains a challenging task for computer vision algorithms with traditional image matching methods. In this paper, we propose a memory segment matching network for image geo-localization that is inspired by the discovery of the place cell in the brain by using artificial intelligence. The place cell becomes active when an animal enters a particular location, where the external sensory information in the environment matches features stored in the hippocampus. In order to emulate the operation of the place cell, we employ a convolutional neural network (CNN) and a long-short term memory (LSTM) to extract the visual features of the environment. The extracted features are stored as segmented memory bounded with a location tag. A matching network is utilized to calculate the cross firing probability of the memory segment and the current input visual data. The final prediction of the location is obtained by sending the cross firing probability to an inference engine that uses a hidden Markov model (HMM). According to the simulation results, the localization accuracy reaches up to 95% for the datasets tested, which outperforms the state-of-the-art by 17% in localization detection accuracy.

Radio frequency identification (RFID) is widely used in various areas such as logistics, supply-chain management, and access control. The new challenge for designing an RFID tag is how to embed a low-cost and low-power consumption algorithm into a compact RFID tag chip. This paper presents a passive HF-band tag chip supporting ISO/IEC 14443 type A/B protocol with low-cost and low-power consumption. In the Analog Front End, a bridge rectifier with 73.76% power conversion efficiency is presented to accomplish RF-dc conversion. A robust demodulator with both 10% and 100% ASK demodulation capabilities and a subcarrier-based modulator are designed to complete the data transfer process. Moreover, a burr-eliminating power ON/OFF reset circuit is proposed to provide a reset signal for the system. In the digital baseband controller, a linear feedback shift register-based light-weight authentication protocol is presented to ensure data security while reducing resource overhead. The embedded 8-Kb EEPROM contains eight independent keys for eight different application fields. The tag chip is fabricated in HJ025 2P4M CMOS process with an area of 1.1mm $\times $ 1.18mm and total static power consumption of $116.45~\mu \text{W}$ . The low cost and low-power consumption ensure the tag chip, especially suitable for smart cards.

Nowadays video content has contributed to the majority of Internet traffic, which brings great challenge to the network infrastructure. Fortunately, the emergence of edge computing has provided a promising way to reduce the video load on the network by caching contents closer to users.But caching replacement algorithm is essential for the cache efficiency considering the limited cache space under existing edge-assisted network architecture. To investigate the challenges and opportunities inside, we first measure the performance of five state-of-the-art caching algorithms based on three real-world datasets. Our observation shows that state-of-the-art caching replacement algorithms suffer from following weaknesses: 1) the rule-based replacement approachs (e.g., LFU,LRU) cannot adapt under different scenarios; 2) data-driven forecast approaches only work efficiently on specific scenarios or datasets, as the extracted features working on one dataset may not work on another one. Motivated by these observations and edge-assisted computation capacity, we then propose an edge-assisted intelligent caching replacement framework LSTM-C based on deep Long Short-Term Memory network, which contains two types of modules: 1) four basic modules manage the coordination among content requests, content replace, cache space, service management; 2) three learning-based modules enable the online deep learning to provide intelligent caching strategy. Supported by this design, LSTM-C learns the pattern of content popularity at long and short time scales as well as determines the cache replacement policy. Most important, LSTM-C represents the request pattern with built-in memory cells, thus requires no data pre-processing, pre-programmed model or additional information. Our experiment results show that LSTM-C outperforms state-of-the-art methods in cache hit rate on three real-traces of video requests. When the cache size is limited, LSTM-C outperforms baselines by 20%~32% in cache hit rate. We also show that the training and predicting time of one iteration are $8.6~ms$ and $300~\mu s$ on average respectively, which are fast enough for online operations.

In recent years, with the rapid deployment of various Internet of Things (IoT) devices, it becomes a crucial and practical challenge to enable real-time search for objects, data, and services in the Internet of Everything. The IoT data prediction model can not only provide solutions for the real-time acquisition of the IoT sensor data but also provide more meaningful applications than the traditional IoT event detection model. In this paper, we use the complex time series formed by various types of sensors to establish a multi-dimensional feature selection model and a dynamic sensor-data prediction model. Compared with the traditional data prediction model, our model improves the accuracy and stability of the long-term prediction results of the IoT sensor data. Finally, we evaluate our prediction model using the Intel Berkeley Research Lab sensor data with an accuracy of over 98% and 92% accuracy on the Chicago Park District weather&water data.

The rule-based fuzzy systems have successfully applied for numerous medical data classification problems. However, structuring the concise and interpretable fuzzy rules with good classification performance is still a big challenge. To address this issue, a novel feature selection and rule generation integrated learning for Takagi-Sugeno-Kang fuzzy system (called FSRG-IL-TSK) in this paper. FSRG-IL-TSK represents feature selection, structure identification and parameter learning into a Bayesian model, and uses the sequential importance resampling (SIR) algorithm to obtain the optimal parameters simultaneously, including the optimal features for each fuzzy rule, number of rules, and antecedent/consequent parameter of rules. Due to an integrated learning mechanism, it can select a small set of useful features and obtain a small number of rules. The effectiveness and advantages of FSRG-IL-TSK are validated experimentally on real-world medical data classification tasks.

Condition-based maintenance is an effective method for deciding when to maintain deteriorating system. In this paper, a non-periodic maintenance model with adaptive inspection intervals is proposed by the consideration of system stability and deterioration. Different maintenance plans and monitoring strategies are adopted in distinct stages of life cycle of the deteriorating system to reduce costs, which is a complex multiple optimization parameter problem. And particle swarm optimization algorithm combining heuristic rules is designed to solve this multi-objective problem. Finally, a numerical example is implemented to illustrate the effectiveness and rationality of the proposed model. The comparison between the proposed model and other maintenance models illustrates the economy of the proposed model and the importance of considering system stability. Sensitivity analysis is also performed to investigate the effect of seven cost factors.