The research on detecting violent behavior in videos has made good progress, which provides support for monitoring abnormal spread the network, so as to achieve effect of purifying network space environment. A large number current violence detection models have achieved performance experimental environments, but their generalization ability is insufficient. Violent often occurs a variety scenarios, automatic requires model with strong generalization. In this paper, crowd based human contour...

Baseline estimation is a critical component for latent factor-based collaborative filtering (CF) recommendations to obtain baseline predictions by evaluating global deviations both users and items from personalized ratings. Classical presupposes that the user’s factual rating range same as system’s given range. However, observations on real datasets of movie recommender systems, we found different have actual ranges, can be classified into four kinds according their criterion, including...

Multi-source diffusion is a common phenomenon in online social networks (OSNs) that involves pieces of information with different purposes concurrently propagating over cooperative or competitive way. A traditional activated probability model for single-source cannot be directly applied to multi-source diffusion, especially when these diffusions interfere each other. We consider herd behavior exists OSNs, and we propose more rational pattern (APMSID), inspired by the herding effect. APMSID...

The authors present an approach to isolated-word speech recognition which is characterized by two aspects: (1) nonlinear time normalization based on the gradients of short-time energy in a specific number frequency bands, retains transient portions and ignores steady-state signal domain; (2) real-time implementation due low computational load. Simulation has shown that correct rate was 99.5% for multiple speakers TI-20 database. A very high accuracy on-line also obtained.< <ETX...

Social networks have been part of human beings' daily lives and affect nearly every aspect our lives. influence prediction is an interesting topic to predict whether users will or not be activated by current social spreading events, deep learning-based approaches can obtain outstanding accuracy graph neural (GNNs). However, GNN models are restricted the 1-Weisfeiler-Lehman (WL) test represent node structure only first layer neighbors but cannot discriminate nodes with different second more...

This paper analyzes the limitation of baseline estimation by defining four kinds rating personalization corresponding to users' criterions, including Normal, Strict, Lenient, and Middle. We find a standard deviation proportion pattern from ratings' normal distribution enhance handling capability personalized behavior, propose novel model based on Standard Deviation Proportion, named SDP model, improve accuracy existing recommendation algorithms which used traditional estimation. also two...

In recent multi-speaker speech separation researches, the overall deep-learning-based architecture consists of three parts: encoder, separator, and decoder. But improvement strategies generally only focus on separator in middle, regardless its input. The most common encoder structure at present is a single 1D convolution layer followed by nonlinear activation function, ReLU. this paper, we firstly propose new named Attention DE, trying to improve input effectiveness separator. adds extra...

Speaker recognition using i-vector has been replaced by speaker deep learning. based on Convolutional Neural Networks (CNNs) widely used in recent years, which learn low-level speech representations from raw waveforms. On this basis, a CNN architecture called SincNet proposes kind of unique convolutional layer, achieved band-pass filters. Compared with standard CNNs, learns the low and high cut-off frequencies each filter. This paper an improved CNNs PF-Net, encourages first layer to...

Recently rumors have been rapidly propagated while the Internet has extensively developed. Research shows that highly credible comments with a distinct stance worthy information. In this paper, we attempt to combine user credibility and capture during information-dissemination process detect rumors. We propose User Stance Bi-Directional Graph Attention Networks (USBGAT) model extract accurate information for rumor detection based on high users strong stance, diminish ineffectively neutral...

The fault diagnosis of rolling bearings is a critical technique to realize predictive maintenance for mechanical condition monitoring. In real industrial systems, the main challenges pertain accuracy and real-time requirements. Most existing methods focus on ensuring accuracy, requirement often neglected. this paper, considering both requirements, we propose novel fast method bearings, based extreme learning machine (ELM) logistic mapping, named logistic-ELM. First, identify 14 kinds...

In this article, we describe a new neural network structure and corresponding sequential training technique for speech recognition. The proposed system is modification of the original time delay (TDNN) Waibel (1989). consists group sub-nets, each isolated word to be recognized corresponds at least one sub-net. Since sub-net deals with only word, it may trained independently. Each TDNN which train algorithm. has attained close 100% accuracy two-speaker, recognition task.< <ETX...