Search

Metastasis, as one of major challenges in the cancer treatment, is responsible for the high mortality of breast cancer. It has been reported that breast cancer cell invasion and metastasis are related to aquaporin 3 (AQP3), which is the transmembrane transport channel for H

The convolutional neural network achieves excellent semantic segmentation results in artificially annotated datasets with complex scenes. However, semantic segmentation methods still suffer from several problems such as low use rate of the features, high computational complexity, and being far from practical real-time application, which bring about challenges for the image semantic segmentation. Two factors are very critical to semantic segmentation task: global context and multilevel semantics. However, generating these two factors will always lead to high complexity. In order to solve this, we propose a novel structure, dual attention fusion module (DAFM), by eliminating structural redundancy. Unlike most of the existing algorithms, we combine the attention mechanism with the depth pyramid pool module (DPPM) to extract accurate dense features for pixel labeling rather than complex expansion convolution. Specifically, we introduce a DPPM to execute the spatial pyramid structure in output and combine the global pool method. The DAFM is introduced in each decoder layer. Finally, the low-level features and high-level features are fused to obtain semantic segmentation result. The experiments and visualization results on Cityscapes and CamVid datasets show that, in real-time semantic segmentation, we have achieved a satisfactory balance between accuracy and speed, which proves the effectiveness of the proposed algorithm. In particular, on a single 1080ti GPU computer, ResNet-18 produces 75.53% MIoU at 70 FPS on Cityscapes and 73.96% MIoU at 109 FPS on CamVid.

As people pay more and more attention to medicine safety issues, related medicine safety monitoring platforms are also rapidly popularized. However, previous work has poor accuracy and low efficiency in medicine safety assessment. In this paper, the medicine safety evaluation index system of the medicine safety monitoring platform is determined from four aspects: medicine research and development, medicine market, medicine production, and medicine uses. In order to solve the problems of the medicine safety evaluation model, such as low evaluation accuracy, slow convergence speed, and long training time, the dynamic dual optimization of PSO-BP medicine safety assessment method (OPSO-BP) is proposed. The weights and thresholds of BP neural network are optimized by the PSO algorithm to improve the quality of assessment. In addition, we optimize PSO: use the cosine function to dynamically adjust the inertia weight w and use the average optimal position of the individual in the population to replace the optimal position of the individual. It improves the problem that the evaluation model in the traditional algorithm is easy to fall into the local optimal solution due to the lack of generalization ability. In this paper, the effectiveness of OPSO-BP is verified by comparative experiments with the designed questionnaire data of medicine safety evaluation. &nbsp

BackgroundPerforming biopsy for intermediate lesions with PI-RADS 3 has always been controversial. Moreover, it is difficult to differentiate prostate cancer (PCa) and benign prostatic hyperplasia (BPH) nodules in PI-RADS 3 lesions by conventional scans, especially for transition zone (TZ) lesions. The purpose of this study is sub-differentiation of transition zone (TZ) PI-RADS 3 lesions using intravoxel incoherent motion (IVIM), stretched exponential model, and diffusion kurtosis imaging (DKI) to aid the biopsy decision process.MethodsA total of 198 TZ PI-RADS 3 lesions were included. 149 lesions were BPH, while 49 lesions were PCa, including 37 non-clinical significant PCa (non-csPCa) lesions and 12 clinical significant PCa (csPCa) lesions. Binary logistic regression analysis was used to examine which parameters could predict PCa in TZ PI-RADS 3 lesions. The ROC curve was used to test diagnostic efficiency in distinguishing PCa from TZ PI-RADS 3 lesions, while one-way ANOVA analysis was used to examine which parameters were statistically significant among BPH, non-csPCa and csPCa.ResultsThe logistic model was statistically significant (χ2 = 181.410, pp=0.004), mean diffusion (MD) (p=0.005), mean kurtosis (MK) (p=0.015), diffusion coefficient (D) (p=0.001), and distribute diffusion coefficient (DDC) (p=0.038) were statistically significant in the model. ROC analysis showed that AUC was 0.9197 (CI 95%: 0.8736-0.9659). Sensitivity, specificity, positive predictive value and negative predictive value were 92.1%, 80.4%, 93.9% and 75.5%, respectively. FA and MK of csPCa were higher than those of non-csPCa (all ppConclusionFA, MD, MK, D, and DDC can predict PCa in TZ PI-RADS 3 lesions and inform the decision-making process of whether or not to perform a biopsy. Moreover, FA, MD, MK, D, DDC, and ADC may have ability to identify csPCa and non-csPCa in TZ PI-RADS 3 lesions.

The off-target frequency of CRISPR/Cas9 system can be significantly decreased via zwitterionic polymer conjugation or (EK)n peptide fusion while maintaining a similar level of on-target gene editing activity.

Chinese couplets, as one of the traditional Chinese culture, is the treasure of Chinese civilization and the inheritance of Chinese history. Given a sentence (namely an antecedent clause), people reply with another sentence (namely a subsequent clause) equal in length. Because of the complexity of the semantic and grammatical rules of couplet, it is not easy to create a suitable couplet that meets the requirements of sentence pattern, context, and flatness. With the development of neural models and natural language processing, automatic generation of Chinese couplets has drawn significant attention due to its artistic and cultural value, most of these works mainly focus on generating couplet by given text information, while visual inspirations for couplet generation have been rarely explored. In this paper, we design a Chinese couplet generation model based on NIC (Neural Image Caption), which can compose a piece of couplet suitable to the artistic conception in an image. At first, we use the improved VGG16 model to predict the input image. The content of the image can be automatically recognized and the corresponding description are generated and translated into Chinese keywords. Then, the encoder-decoder framework is used repeatedly to process these keywords, and finally the couplet can be generated. Moreover, to satisfy special characteristics of couplets, we incorporate the attention mechanism into the encoding-decoding process, which greatly improves the accuracy of couplets generated automatically.

Automatically generating captions for videos faces a huge challenge since it is a cross-modal cross task that involves vision and texts. Most of the existing models generate the captioning words merely based on the video visual content features, ignoring the important underlying semantic information. The relationship between explicit semantics and hidden visual content is not holistically exploited, thus hardly describing fine-grained caption accurately from a global view. To better extract and integrate the semantic information, we propose a novel encoder-decoder framework of bi-directional long short-term memory with attention model and conversion gate (BiLSTM-CG), which transfers auxiliary attributes and then generates detailed captioning. Specifically, we extract semantic attributes from sliced frames in a multiple-instance learning (MIL) manner. MIL algorithms attempt to learn a classification function that can predict the labels of bags and/or instances in the visual content. In the encoding stage, we adopt 2D and 3D convolutional neural networks to encode video clips, and then feed the concatenate features into a BiLSTM. In decoding stage, we design a CG to adaptively fuse semantic attributes into hidden features at word level, and a CG can convert auxiliary attributes and textual embedding for video captioning. Furthermore, the CG has an ability to automatically decide the optimal time stamp to capture the explicit semantic or rely on the hidden states of the language model to generate the next word. Extensive experiments conducted on the MSR-VTT and MSVD video captioning datasets demonstrate the effectiveness of our method compared with state-of-the-art approaches.

A staged adaptive firefly algorithm (SAFA) is proposed in this paper. Firstly, the attraction model is improved to promote the convergence of the algorithm in the case of small algorithm complexity. Secondly, three adaptive adjustment functions of parameters are established according to the actual conditions of convergence and iteration. Because of the new attraction model, SAFA has better population diversity at the early stage of iteration and can carry out adaptive balance and adjustment of global and local optimization at the late stage of iteration. Because of three adaptive adjustment functions of parameters, SAFA has better randomness and non-repeatability of parameters, so it has stronger global convergence ability. To verify the performance, SAFA algorithm is compared with other four algorithms in testing six standard functions and unmanned aerial vehicle (UAV) charging path planning for wireless sensor network in this paper. A large number of experimental results show that the precision and convergence speed of SAFA is higher than that of the other four algorithms.

Despite the great progress achieved, image-to-video person re-identification is still challenging in the cross-modal scenario. Currently, state-of-the-art approaches mainly concentrate on the task-specific data, neglecting the extra information from the different but related tasks. In this paper, we propose an end-to-end neural network framework for image-to-video person re-identification with cross-modal embeddings learned from extra information. Concretely speaking, cross-modal embedding layers from image captioning and video captioning models, are incorporated to learn common latent embeddings for multiple modalities. The learned multimodal embeddings are expected to focus on person’s prominent distinctions, due to textual descriptive information generally paying close attention to person’s explicit characteristics. Apart from that, our proposed framework resorts to CNNs and LSTMs for extracting visual and spatiotemporal features, and combines the strengths of identification and verification model to improve the discriminative ability of the learned features. The experimental results demonstrate the effectiveness of our framework on narrowing down the gap between heterogeneous data and obtaining observable improvement in the image-to-video person re-identification task.

AimsOur objective is to identify first-trimester plasmatic miRNAs associated with and predictive of GDM.MethodsWe quantified miRNA using next-generation sequencing in discovery (Gen3G: n = 443/GDM = 56) and replication (3D: n = 139/GDM = 76) cohorts. We have diagnosed GDM using a 75-g oral glucose tolerance test and the IADPSG criteria. We applied stepwise logistic regression analysis among replicated miRNAs to build prediction models.ResultsWe identified 17 miRNAs associated with GDM development in both cohorts. The prediction performance of hsa-miR-517a-3p|hsa-miR-517b-3p, hsa-miR-218-5p, and hsa-let7a-3p was slightly better than GDM classic risk factors (age, BMI, familial history of type 2 diabetes, history of GDM or macrosomia, and HbA1c) (AUC 0.78 vs. 0.75). MiRNAs and GDM classic risk factors together further improved the prediction values [AUC 0.84 (95% CI 0.73–0.94)]. These results were replicated in 3D, although weaker predictive values were obtained. We suggest very low and higher risk GDM thresholds, which could be used to identify women who could do without a diagnostic test for GDM and women most likely to benefit from an early GDM prevention program.ConclusionsIn summary, three miRNAs combined with classic GDM risk factors provide excellent prediction values, potentially strong enough to improve early detection and prevention of GDM.

Data from: Maternal thyroid dysfunction and neurodevelopment in children.

An essential device for managing the vectoring nozzle’s thrust direction is the thrust-vectoring nozzle adjusting mechanism. The attitude of an aircraft’s flight may be changed by adjusting the vectoring nozzle’s airflow, which affects the direction of thrust. The adjusting mechanism’s motion precision and operational stability are crucial in this procedure. In this paper, the influence of the clearance of the motion joint on the dynamic characteristics of the thrust-vectoring nozzle adjusting mechanism under different flight conditions are studied. First, a dynamic model of the thrust-vectoring nozzle adjusting mechanism that takes into account the clearance and friction properties of the motion joint is constructed by combining the Lagrange equation of the first sort with the mixed contact force model and the LuGre friction model. Second, the variation law of nonlinear contact stiffness is explored, and the precision of the dynamic model is confirmed by contrasting the results of the numerical simulation with those of the Simulink simulation. Finally, the simulation results reveal that while the motion joint’s clearance causes the adjusting blade’s velocity and acceleration to vibrate violently and momentarily during early operation, this clearance has minimal effect on the precision of blade motion throughout the stable phase.

This paper presents a three-tier modality alignment approach to learning text-image joint embedding, coined as JEMA, for cross-modal retrieval of cooking recipes and food images. The first tier improves recipe text embedding by optimizing the LSTM networks with term extraction and ranking enhanced sequence patterns, and optimizes the image embedding by combining the ResNeXt-101 image encoder with the category embedding using wideResNet-50 with word2vec. The second tier modality alignment optimizes the textual-visual joint embedding loss function using a double batch-hard triplet loss with soft-margin optimization. The third modality alignment incorporates two types of cross-modality alignments as the auxiliary loss regularizations to further reduce the alignment errors in the joint learning of the two modality-specific embedding functions. The category-based cross-modal alignment aims to align the image category with the recipe category as a loss regularization to the joint embedding. The cross-modal discriminator-based alignment aims to add the visual-textual embedding distribution alignment to further regularize the joint embedding loss. Extensive experiments with the one-million recipes benchmark dataset Recipe1M demonstrate that the proposed JEMA approach outperforms the state-of-the-art cross-modal embedding methods for both image-to-recipe and recipe-to-image retrievals., accepted by CIKM 2021. arXiv admin note: substantial text overlap with arXiv:2108.00705

Recently, the Capsule Network is an emerging neural network structure that is characterized by the ability to maintain high classification accuracy. By analyzing the difference between Capsule Network and traditional convolutional neural network, it is found that the model compression method applied to the traditional neural network cannot be directly used in the Capsule Network. To address the problem, an IPC-CapsNet compression algorithm is proposed based on the structural characteristics of the Capsule Networks. The algorithm can reduce the computational complexity and compress the scale of model computation on the basis of retaining the accuracy of model classification. Considering the deficiency of Capsule Network processing serialized text data separately, we combined with IPC-CapsNet and then come up with a sentiment classification algorithm SPT-CapsNet. It has conducted a sentiment analysis experiment of MicroBlog dataset. Compared to other methods, our SPT-CapsNet obtained the best performance among the metrics. The SPT-CapsNet improves the running speed and maintains the balance between classification accuracy and computational efficiency.

An extract prepared from species of Paris is the most widely consumed herbal product in China. The genus Paris includes a variety of genotypes with different medicinal component contents but only two are defined as official sources. Closely related species have different medicinal properties because of differential expression of proteins and metabolites. To better understand the molecular basis of these differences, we examined proteomic and metabolomic changes in rhizomes of P. polyphylla var. chinensis, P. polyphylla var. yunnanensis, and P. fargesii var. fargesii using a technique known as sequential window acquisition of all theoretical mass spectra as well as gas chromatography-time-of-flight mass spectrometry. In total, 419 proteins showed significant abundance changes, and 33 metabolites could be used to discriminate Paris species. A complex analysis of proteomic and metabolomic data revealed a higher efficiency of sucrose utilization and an elevated protein abundance in the sugar metabolic pathway of P. polyphylla var. chinensis. The pyruvate content and efficiency of acetyl-CoA-utilization in saponin biosynthesis were also higher in P. polyphylla var. chinensis than in the other two species. The results expand our understanding of the proteome and metabolome of Paris and offer new insights into the species-specific traits of these herbaceous plants. SIGNIFICANCE: The traditional Chinese medicine Paris is the most widely consumed herbal product for the treatment of joint pain, rheumatoid arthritis and antineoplastic. All Paris species have roughly the same morphological characteristics; however, different members have different medicinal compound contents. Efficient exploitation of genetic diversity is a key factor in the development of rare medicinal plants with improved agronomic traits and malleability to challenging environmental conditions. Nevertheless, only a partial understanding of physiological and molecular mechanisms of different plants of Paris can be achieved without proteomics. To better understand the molecular basis of these differences and facilitate the use of other Paris species, we examine proteomic metabolomic changes in rhizomes of Paris using the technique known as SWATH-MS and GC/TOF-MS. Our research has provided information that can be used in other studies to compare metabolic traits in different Paris species. Our findings can also serve as a theoretical basis for the selection and cultivation of other Paris species with a higher medicinal value.

With the rapid development of deep neural networks, multimedia data has been greatly enriched during the past few years. Searching among multimedia data with a variety of modalities (i.e., cross-modal retrieval) has become a central research topic due to its ability to better express semantics compared to simple uni-modal search. These methods give amazing results in many areas, but it is difficult to deal with many objects and relationships faithfully reproducing complex sentences. In this paper, we propose a novel approach for image-sentence retrieval to overcome this limitation. Our approach leverages the semantic expressiveness of the scene graph and incorporates it into a deep neural network model for more efficient retrieval. We demonstrate the effectiveness of this approach in real-world and simulated computer vision tasks. Extensive experiments show that our method can significantly improve the accuracy of the image-sentence retrieval task.

Additional file 1. Supplementary tables.

Human action recognition is an essential area of research in the field of computer vision. However, existing methods ignore the essence of infrared image spectral imaging. Compared with the visible modality with all three channels, the infrared modality with approximate single-channel pays more attention to the lightness contrast and loses the channel information. Therefore, we explore channel duplication and tend to investigate more appropriate feature presentations. We propose a subspace enhancement and colorization network (S\(^2\)ECNet) to recognize infrared video action recognition. Specifically, we apply the subspace enhancement (S\(^2\)E) module to promote edge contour extraction with subspace. Meanwhile, a subspace colorization (S\(^2\)C) module is utilized for better completing missing semantic information. What is more, the optical flow provides effective supplements for temporal information. Experiments conducted on the infrared action recognition dataset InfAR demonstrates the competitiveness of the proposed method compared with the state-of-the-arts.

Degradation of tetracycline hydrochloride (TCH) was carried out using natural light driven BiOI photocatalyst through sonophotocatalytic technique. Sonocatalytic degradation was considered as a new and advanced strategy for the elimination of hazardous organic pollutants from wastewater. Until now, the synergy of employing ultrasonic irradiation with BiOI based on catalysts for the removal of antibiotic residues is rarely reported. BiOI was an excellent semiconductor catalyst of Bi class. In this paper, mesoporous BiOI microspheres self-assembled from nanosheets were prepared by solvothermal method and its ultrasonic catalytic performance was studied. The sonocatalytic performance of BiOI was evaluated in terms of the degradation of tetracycline hydrochloride (TCH) as a simulate pollutant under ultrasonic irradiation. Three comparative experiments were designed to explore the degradation of TCH in natural light as follows: ultrasound alone, BiOI alone and ultrasound/BiOI synergy. The degradation rate of TCH by ultrasound/BiOI synergy was 227 times higher than that catalyzed by ultrasound alone and 83 times higher than that catalyzed by BiOI, and the maximum degradation ratio of TCH could reach to 93.0%. Therefore, the synergistic effect was significant, and the value of synergy factor was estimated as 61. Many factors, such as the ultrasonic power, catalyst concentration (Ccatalytic), ultrasonic duty cycle, concentration of initial TCH dye (CTCH), could affect the ultrasonic degradation efficiency. The BBD methodology with RSM was applied for modeling, optimization and investigation of influence of operational parameters, i.e. ultrasonic power, catalyst concentration, ultrasonic duty cycle. Analysis of variance (ANOVA) confirmed a good quadratic response surface model for predicting the sonocatalytic efficiency at various operational parameters (R2= 0.9936 and Adjusted R2=0.9854. The possible mechanism of the high degradation rate of TCH might be associated with the generation of reactive oxygen species (ROS) which could improve the chemistry yields.

With the great development of 5G technology research and the constant development of the network shopping, clothing classification and clothing collocation recommendation based on clothing pictures can provide advices to the customer and help businesses to promote sales. Deep learning is a latest research achievement in the field of machine learning, it has a strong ability of image modeling and image representation, which makes breakthrough progress in the field of image processing. Based on image data of dressing commodities provided by Taobao.com, as well as the text data of both customers' historical behaviors and dressing outfits generated by fashion experts, we design and implement clothing collocation and recommendation through relevant technologies of data mining and deep learning.

Textual-visual cross-modal retrieval has been an active research area in both computer vision and natural language processing communities. Most existing works learn a joint embedding model that maps raw text-image pairs onto a joint latent representation space in which the similarity between textual embeddings and visual embeddings can be computed and compared, without leveraging diverse semantics. This paper presents a general framework to study and evaluate the impact of diverse semantics extracted from the multi-modal input data on the quality and performance of joint embedding learning. We identify different ways that conventional textual features, such as TFIDF term frequency semantics and image category semantics, can be combined with neural features to further boost the efficiency of joint embedding learning. Experiments on the benchmark dataset Recipe1M demonstrates that existing representative cross-modal joint embedding approaches enhanced with diverse semantics in both raw inputs and joint embedding loss optimization can effectively boost their cross-modal retrieval performance.

Video captioning task is to generate a text to describe the content in the video. To generate a proper description, many people have begun to add explicit semantic information to the video generation process. However, in recent work, with the mining of semantics in video, the semantic information in some existing methods will play a smaller and smaller role in the decoding process. Besides, decoders apply temporal attention mechanisms to all generation words including visual vocabulary and non visual vocabulary that will produce inaccurate or even wrong results. To overcome the limitations, 1) we detect visual feature to composite semantic tags from each video frame and introduce a semantic combination network in the decoding stage. We use the probability of each semantic object as an additional parameter in the long-short term memory(LSTM), so as to better play the role of semantic tags, 2) we combine two levels of LSTM with temporal attention mechanism and adaptive attention mechanism respectively. Then we propose an adaptive attention mechanism based semantic compositional network (AASCNet) for video captioning. Specifically, the framework uses temporal attention mechanism to select specific visual features to predict the next word, and the adaptive attention mechanism to determine whether it depends on visual features or context information. Extensive experiments conducted on the MSVD video captioning dataset prove the effectiveness of our method compared with state-of-the-art approaches.

As we all know, inconsistent distribution and representation of different modalities, such as image, text and audio, cause the “media gap”, which poses a great challenge to deal with such heterogeneous data. Currently, state-of-the-art multimodal approaches mainly focus on the data provided by target task, neglecting the extra information on different but related tasks. In this paper, we explore a multimodal representation learning architecture by leveraging embedding representation trained from extra information. Specifically speaking, the approach of fixed model reuse is integrated into our architecture, which can incorporate helpful information from existing models/features into a new model. Based on our proposed architecture, we study multilingual OCR and long-text-based image retrieval tasks. Multilingual OCR is a difficult task that deals with multiple languages on the same page. We take advantage of extra textual embedding layer in an existing text-generating model to improve the accuracy of multilingual OCR. As for the long-text-based image retrieval, a cross-modal task, intermediate visual embedding layer in an off-the-shelf image-captioning model is leveraged to enhance the retrieval ability. The experimental results validate the effectiveness of our proposed architecture on narrowing down the “media gap” and yield observable improvement in these two tasks. Our architecture outperform the state-of-the-art approaches by 4.2% improvements in terms of accuracy in multilingual OCR task and yields improvement from 9 to 6 with regard to the median rank of retrieval result in long-text-based image retrieval task.

Autonomic Nervous System In article number 2100808 by Yan Zhang, Hui Xie, and co‐workers, osteocyte neuropeptide Y (NPY)‐dependent neuronal control of bone marrow mesenchymal stem/stromal cell (BMSC) fate decision is found. Osteocyte NPY promotes bone marrow adipogenesis at the expense of osteogenesis by BMSC. Osteocyte NPY production is controlled by autonomic nervous system (ANS). γ‐Oryzanol attenuates ANS dysregulation, NPY overproduction, and bone‐fat imbalance induced by aging and estrogen deficiency. [Image: see text]

Difference system of sets (DSS), introduced by Levenshtein, has an interesting connection with the construction of comma-free codes. In this paper, we construct two new families of DSS from the cyclotomic classes of order 12.

Additional file 1.

With the explosive growth of mobile devices, it is feasible to deploy image recognition applications on mobile devices to provide image recognition services. However, traditional mobile cloud computing architecture cannot meet the demands of real time response and high accuracy since users require to upload raw images to the remote central cloud servers. The emerging architecture, Mobile Edge Computing (MEC) deploys small scale servers at the edge of the network, which can provide computing and storage resources for image recognition applications. To this end, in this paper, we aim to use the MEC architecture to provide image recognition service. Moreover, in order to guarantee the real time response and high accuracy, we also provide a feature extraction algorithm to extract discriminative features from the raw image to improve the accuracy of the image recognition applications. In doing so, the response time can be further reduced and the accuracy can be improved. The experimental results show that the combination between MEC architecture and the proposed feature extraction algorithm not only can greatly reduce the response time, but also improve the accuracy of the image recognition applications.

Hydraulic manipulator are widely used in underwater detection, anti-terrorism, industrial palletizing and other fields. The research object of this paper is the self-developed Hydraulic Rotary Vane Actuator (RVA) as the joint of the six-degree-of-freedom hydraulic manipulator. The kinematics error model of the joint flexible hydraulic manipulator is established by the improved D-H method. The optimal joint angle is determined, and the error distribution in the optimal joint angle and the general joint angle is compared and analyzed. Finally, the rajectory planning is used to reduce the error distribution of End-Effector(EE), vibration and shock.

In this paper, by combing the collected testing data of pile bearing capacity from 78 reinforced concrete cast-in-place bored piles. The distribution characteristics of the pile bearing capacity are analyzed in detail. Based on this, the n-σ criteria are introduced and a more practical data processing method for bearing capacity of foundation piles is proposed. Using this method, the data of pile bearing capacity detected was analyzed and processed. Then the data was divided into "strong data", "good data" and "weak data". In addition, we verified the validity of this method to determine the detected data quality of pile bearing capacity through engineering examples. The verification shows that the data quality has a significant influence on the calculation results of the reliability index and the resistance coefficient.

A new improved CSMA/CA algorithm for wireless sensor networks (WSNs) is proposed in this paper to save energy and prolong the life cycle of WSNs. This algorithm is combined with the artificial neural network and Bayesian algorithm according to the practical applications of the drug control system of the Internet of Things. The algorithm is divided into two parts: first, the artificial neural network algorithm is used to estimate the data of WSNs, the results are the reference for the conversion of routing node frequency; second, by using the Bayesian formula, valuation method, and the CSMA/CA’s collision detection mechanism, the algorithm adjusts the frequency of the routing node and the relevant node frequency to establish the normal communication of packets sent by nodes and the aggregation node packets. In this way, it will reduce the collision detection and the back off time and avoid data packet duplication. The simulation tool-NS2 is used to configure an appropriate simulation scene for the experiment, which analyses and compares the received packet rate, the overall energy consumption of the network, and so on. The results demonstrate that the proposed algorithm ensures high energy efficiency and balanced energy consumption. Therefore the results show that the improved algorithm increases the efficiency, so that the network has the function of intelligent learning.

To solve the distributed task allocation problems of search and rescue missions for multiple unmanned aerial vehicles (UAVs), this paper establishes a dynamic task allocation model under three conditions: 1) when new targets are detected, 2) when UAVs break down and 3) when unexpected threats suddenly occur. A distributed immune multi-agent algorithm (DIMAA) based on an immune multi-agent network framework is then proposed. The technologies employed by the proposed algorithm include a multi-agent system (MAS) with immune memory, neighbourhood clonal selection, neighbourhood suppression, neighbourhood crossover and self-learning operators. The DIMAA algorithm simplifies the decision-making process among agents. The simulation results show that this algorithm not only obtains the global optimum solution, but also reduces the communication load between agents.

Chinese couplets, as one of the traditional Chinese culture, is the treasure of Chinese civilization and the inheritance of Chinese history. Given a sentence (namely an antecedent clause), people reply with another sentence (namely a subsequent clause) equal in length. Because of the complexity of the semantic and grammatical rules of couplet, it is not easy to create a suitable couplet that meets the requirements of sentence pattern, context, and flatness. In this paper, given the issued antecedent clause, we can automatically generate the subsequent clause by encoder-decoder model. Moreover, to satisfy special characteristics of couplets, we incorporate the attention mechanism into the encoding-decoding process, which greatly improves the accuracy of couplets generated automatically.

User profiling means exploiting the technology of machine learning to predict attributes of users, such as demographic attributes, hobby attributes, preference attributes, etc. It’s a powerful data support of precision marketing. Existing methods mainly study network behavior, personal preferences and post texts to build user profile. Through our data analysis of micro-blog, we find that females show more positive and have richer sentiments than males in online social platform. This difference is very conducive to the distinction between genders. Therefore, we argue that sentiment context is important as well for user profiling. In this paper, we propose to predict one of the demographic labels: gender by exploring micro-blog user posts. Firstly we build a sentiment polarity classifier in advance by training a Long Short-Term Memory (LSTM) model. Next we extract sentiment representations from LSTM middle layer. Lastly we combine sentiment representations with virtual document vectors to train a basic MLP network for gender classification. We conduct experiments on a dataset provided by SMP CUP 2016 in China. Experimental results show that our approach can improve gender classification accuracy by 5.53%, compared with classical MLP gender classification without sentiment context.

Pseudorandom binary sequences with optimal balance and autocorrelation have many applications in stream cipher, communication, coding theory, etc. Constructing a binary sequences with three-level autocorrelation is equivalent to find the corresponding characteristic set of the sequences that should be an almost difference set. In this letter, we construct three new families of almost difference sets and two new families of difference sets from the cyclotomic classes of order twelve.

With its continuous development, location information acquisition technology is able to collect more and more trajectory data, and the rich information contained therein is gradually attracting attention from researchers. Trajectory data involves complex relationships among moving objects, time, space, which are hard to understand and be used directly. Nowadays, visual analysis of trajectory data is mainly focus on its representation and interaction, but fails to address the complex correlation contained in trajectory data. Hence, we propose TrajHIN, a heterogeneous information network model built on trajectory data, measure the meta path-based similarity and centrality, and use a visual analytics method to deeply understand trajectory data. The example of visual analysis of real trajectory data has been interpreted and given feedback from domain experts, which proves effectiveness of TrajHIN and feasibility of mining implicit semantic information from trajectory data.

Nowadays, there is a great demand for multilingual optical character recognition (MOCR) in various web applications. And recently, Long Short-Term Memory (LSTM) networks have yielded excellent results on Latin-based printed recognition. However, it is not flexible enough to cope with challenges posed by web applications where we need to quickly get an OCR model for a certain set of languages. This paper proposes a Hybrid Model Reuse (HMR) training approach for multilingual OCR task, based on 1D bidirectional LSTM networks coupled with a model reuse scheme. Specifically, Fixed Model Reuse (FMR) scheme is analyzed and incorporated into our approach, which implicitly grabs the useful discriminative information from a fixed text generating model. Moreover, LSTM layers from pre-trained networks for unilingual OCR task are reused to initialize the weights of target networks. Experimental results show that our proposed HMR approach, without assistance of any post-processing techniques, is able to effectively accelerate the training process and finally yield higher accuracy than traditional approaches.