Abstract Background Currently, cardiovascular disease has become a major endangering human health, and the number of such patients is growing. Electrocardiogram (ECG) an important basis for {medical doctors to diagnose disease, which can truly reflect health heart. In this context, contradiction between lack medical resources surge in increasingly prominent. The use computer-aided diagnosis particularly important, so study ECG automatic classification method strong practical significance....

Protein secondary structure prediction is one of the most important and challenging problems in bioinformatics. Machine learning techniques have been applied to solve problem gained substantial success this research area. However there still room for improvement toward theoretical limit. In paper, we present a novel method protein based on data partition semi-random subspace (PSRSM). Data partitioning an strategy our method. First, training dataset was partitioned into several subsets length...

The ReLU activation function accelerates the convergence of training process in classical framework deep learning. causes a large part network neurons to die. When very gradient flows through neuron and updates parameters, it will not activate any data. This paper proposes target recognition based on CNN with LeakyReLU PReLU functions. According advantages ReLU, is used fix parameters cope death. combined are trained construct new framework. Experimental results show method effective feasibile.

Atrial fibrillation is the most common persistent form of arrhythmia. A method based on wavelet transform combined with deep convolutional neural network applied for automatic classification electrocardiograms. Sinc... | Find, read and cite all research you need Tech Science Press

Abstract During music recommendation scenarios, sparsity and cold start problems are inevitable. Auxiliary information has been utilized in algorithms to provide users with more accurate results. This study proposes an end-to-end framework, MMSS_MKR, that uses a knowledge graph as source of auxiliary serve the obtained from it module. The framework exploits Cross & Compression Units bridge embedding task modules. We can obtain realistic triple exclude false much possible, because our...

Pneumonia remains the leading infectious cause of death in children under age five, killing about 700,000 each year and affecting 7% world’s population. X-ray images lung become key to diagnosis this disease, skilled doctors a certain degree subjectivity, if use computer-aided medical automatically detect abnormalities, will improve accuracy diagnosis. This research aims introduce deep learning technology based on combination Xception neural network long-term short-term memory (LSTM), which...

Melanoma is a fatal skin disease, and there are many challenging tasks in the detection of melanoma through neural network at this stage. We propose new method for diagnosis based on EfficientNet patch strategy. The has three stages operation. First, Cyclegan applied offline to synthesize under-represented category samples from over-represented samples, original image synthesized combined into training set complete conditional synthesis task; second, we creatively strategy implement...

With the rapid development of deep learning, use object detection algorithms for aerial insulator image defect has become main way. To address problems low accuracy small targets, weak representation ability feature maps, insufficient extracted key information, and shortage datasets, this paper proposes an improved method named BS-YOLOv5s based on 3-D attention mechanism Bi-Slim-neck using YOLOv5s as base network. Additionally, to solve problem a new dataset Weather-Insulator (WI) containing...

Prediction of protein secondary structure is important to analyze the folding patterns. We propose a prediction method based on support vector machine (SVM) with position-specific scoring matrix (PSSM) profiles in this paper. The PSSM are obtained from CB513 data set and PSI-BLAST program. arrange sliding window 13 dimension feature 260. grid search algorithm genetic used optimize c γ parameters SVM. experimental results show that paper more effective than traditional which using amino acid...

This paper summarizes the classification, implementation and evaluation of question answering system (QA). QA is divided into four categories: chat robot, based knowledge base, retrieval on free text. Web composed with analysis module, information module answer extraction module. In last, TREC outlined.

The highest three-state prediction accuracy of protein secondary structure is now at 82-84% without using templates, approaching to the theoretical limit 88-90%. Increasingly larger training datasets cover more sequences and structures. More powerful deep learning techniques are not only able deal with computation load large data, but also can capture long-range interactions sequence. In this research, we propose a new approach design two dimensional convolutional neural networks (2DCNN) 6...

In computer vision, edge detection is a hot research area in which Canny operator typical algorithm. has preferable anti-noise ability. However the based on not consecutive. GVF Snake model used widely image segmentation. But there are problems convergence processing to boundaries of some medical because noise. This paper presents new segmentation algorithm image. First, rough got by operator, and then thinning method mathematical morphology adopted get map as foundation model. solves...

In this paper we propose a approach using 2D convolutional neural network(CNN) for the prediction of protein secondary structure. A representation two dimensional PSSM is directly used to convolve with filters extract features. This new reflects not only evolutionary information but also sequence interaction residue. The architecture network has layers and one max-pooling layer. feature maps extracted from second layer are feed Bayes classifier, in order build model. Q3 accuracy 77.7% 25PDB...

In recommendation algorithms, data sparsity and cold start problems are inevitable.To solve such problems, researchers apply auxiliary information to mine users' historical records obtain more potential information, then improve performance.In this paper, ST_RippleNet, a model that combines knowledge graphs with deep learning, is proposed.This starts by building the required graph.Then, interest of users mined through graph, which stimulates spread preferences on set entities.In preference...

Medical i mage segmentation is a n important step in 3-D reconstruction , and from medical images an application of computer graphics biomedicine image processing. A improved method which suitable for presented this paper. algorithm used to reconstruct the model images. R ough edge obtained by multi-scale wavelet transform at f irst. With rough edge, gradient vector flow s nake object contour found. In experiment we kidney liver brain putamen . The performances experiments indicate that new...

In this paper we propose an approach to use wavelets and 2D convolutional neural network (CNN) extract features for the prediction of protein secondary structure. A wavelet feature matrix extracted from PSSM profiles is input into features. Wavelets changing information evolutionary matrix, networks catch sequence interaction residue. The maps last layer are used feed Bayes classifier, in order build model. Q3 accuracy 76.9% ASTRAL dataset achieved based on 3 fold cross validation...

Infrared and visible image fusion aims to integrate salient targets abundant texture information into a single fused image. Existing methods typically ignore the issue of illumination, so that there are problems weak details poor visual perception in case low illumination. To address this issue, we propose low-light oriented infrared network, named L2Fusion. In particular, first design decomposition network according Retinex theory obtain reflectance features with low-light. Then, these...