Learning group interaction for sports video understanding from a perspective of athlete

Learning activities interactions between small groups is a key step in understanding team sports videos. Recent research focusing on team sports videos can be strictly regarded from the perspective of the audience rather than the athlete. For team sports videos such as volleyball and basketball videos, there are plenty of intra-team and inter-team relations. In this paper, a new task named Group Scene Graph Generation is introduced to better understand intra-team relations and inter-team relations in sports videos. To tackle this problem, a novel Hierarchical Relation Network is proposed. After all players in a video are finely divided into two teams, the feature of the two teams’ activities and interactions will be enhanced by Graph Convolutional Networks, which are finally recognized to generate Group Scene Graph. For evaluation, built on Volleyball dataset with additional 9660 team activity labels, a Volleyball+ dataset is proposed. A baseline is set for better comparison and our experimental results demonstrate the effectiveness of our method. Moreover, the idea of our method can be directly utilized in another video-based task, Group Activity Recognition. Experiments show the priority of our method and display the link between the two tasks. Finally, from the athlete’s view, we elaborately present an interpretation that shows how to utilize Group Scene Graph to analyze teams’ activities and provide professional gaming suggestions.     

基于知识图谱嵌入与多神经网络的序列推荐算法

循环神经网络在序列推荐中占有重要地位,但在推荐中,用户的行为序列远比自然语言处理中的句子或计算机视觉中的图像要复杂得多。单一的循环神经网络结构难以充分地挖掘用户偏好,因此提出一种新型的序列推荐算法,同时考虑序列的时间信息以及内容信息。主要分为2个部分：改进的项目嵌入和序列偏好学习。首先,提出一种融合知识图谱的项目嵌入方法,用于生成高质量的项目向量;其次,提出一种卷积神经网络结合长短时记忆神经网络的序列建模方法。更进一步地提出一个基于注意力的框架,动态地结合用户的兴趣点。在公开数据集MovieLens10M上与传统方法以及现有的同类型方法进行了比较。实验结果表明,所提算法在推荐评价指标平均倒数排名MRR@N以及召回率Recall@N上有显著的提升,验证了该算法的有效性。     

融合注意力机制和CNN-GRNN模型的读者情绪预测

针对主流面向文本的读者情绪预测算法难以捕捉文本中复杂的语义和语法信息,以及局限于使用多标签分类方法的问题,提出一种融合注意力机制和卷积门限循环神经网络的读者情绪预测方法。该方法将文本划分为多个句子,利用卷积神经网络从每个句子中提取不同粒度的n-gram信息,构建句子级别的特征表示;然后通过门限循环神经网络顺序地集成这些句子特征,并利用注意力机制自适应地感知上下文信息提取影响读者情绪的文本特征;最后利用softmax回归进行细粒度的读者情绪分布预测。在雅虎新闻读者情感分析数据集上的实验结果证明了该方法的有效性。     

CoGCN: Combining co‐attention with graph convolutional network for entity linking with knowledge graphs

Entity linking is a fundamental task in natural language processing. The task of entity linking with knowledge graphs aims at linking mentions in text to their correct entities in a knowledge graph like DBpedia or YAGO2. Most of existing methods rely on hand‐designed features to model the contexts of mentions and entities, which are sparse and hard to calibrate. In this paper, we present a neural model that first combines co‐attention mechanism with graph convolutional network for entity linking with knowledge graphs, which extracts features of mentions and entities from their contexts automatically. Specifically, given the context of a mention and one of its candidate entities' context, we introduce the co‐attention mechanism to learn the relatedness between the mention context and the candidate entity context, and build the mention representation in consideration of such relatedness. Moreover, we propose a context‐aware graph convolutional network for entity representation, which takes both the graph structure of the candidate entity and its relatedness with the mention context into consideration. Experimental results show that our model consistently outperforms the baseline methods on five widely used datasets.     

基于MFE-BERT与FNNAttention的心理医学知识图谱构建

针对心理医学领域文本段落冗长、数据稀疏、知识散乱且规范性差的问题, 提出一种基于多层级特征抽取能力预训练模型(MFE-BERT)与前向神经网络注意力机制(FNNAttention)的心理医学知识图谱构建方法. MFE-BERT在BERT模型基础上将其内部所有Encoder层特征进行合并输出, 以获取包含更多语义的特征向量, 同时对两复合模型采用FNNAttention机制强化词级关系, 解决长文本段落语义稀释问题. 在自建的心理医学数据集中, 设计MFE-BERT-BiLSTM-FNNAttention-CRF和MFE-BERT-CNN-FNNAttention复合神经网络模型分别进行心理医学实体识别和实体关系抽取, 实体识别F1值达到93.91%, 实体关系抽精确率达到了89.29%, 通过融合文本相似度与语义相似度方法进行实体对齐, 将所整理的数据存储在Neo4j图数据库中, 构建出一个含有3652个实体, 2396条关系的心理医学知识图谱. 实验结果表明, 在MFE-BERT模型与FNNAttention机制的基础上构建心理医学知识图谱切实可行, 提出的改进模型所搭建的心理医学知识图谱可以更好地应用于心理医学信息管理中, 为心理医学数据分析提供参考.     

Neural‐network‐based predictive controller design: An application to temperature control of a plastic injection molding process

This paper presents a neural‐network‐based predictive control (NPC) method for a class of discrete‐time multi‐input multi‐output (MIMO) systems. A discrete‐time mathematical model using a recurrent neural network (RNN) is constructed and a learning algorithm adopting an adaptive learning rate (ALR) approach is employed to identify the unknown parameters in the recurrent neural network model (RNNM). The NPC controller is derived based on a modified predictive performance criterion, and its convergence is guaranteed by adopting an optimal algorithm with an adaptive optimal rate (AOR) approach. The stability analysis of the overall MIMO control system is well proven by the Lyapunov stability theory. A real‐time control algorithm is proposed which has been implemented using a digital signal processor, TMS320C31 from Texas Instruments. Two examples, including the control of a MIMO nonlinear system and the control of a plastic injection molding process, are used to demonstrate the effectiveness of the proposed strategy. Results from both numerical simulations and experiments show that the proposed method is capable of controlling MIMO systems with satisfactory tracking performance under setpoint and load changes. Copyright © 2010 John Wiley and Sons Asia Pte Ltd and Chinese Automatic Control Society     

A hybrid recurrent neural network‐logistic chaos‐based whale optimization framework for heart disease prediction with electronic health records

Heart disease, known interchangeably as “Cardio Vascular Disease,” blocks the blood vessels in the heart and causes heart attack, chest pain, and stroke. Heart disease is one of the leading causes of morbidity and mortality worldwide and it is one of the major causes of morbidity and mortality globally and a trending topic in clinical data analysis. Assessing risk factors related to heart disease is considered as an important step in diagnosing the disease at an early stage. Clinical data present in the form of electronic health records (EHR) can be extracted with the aid of machine learning (ML) algorithms to provide valuable decisions and predictions. ML approaches also play a vital role in early diagnosis and therapeutic monitoring of heart disease. Several research works have been carried out recently to predict heart disease. To this end, we propose a novel hybrid recurrent neural network (RNN)‐logistic chaos‐based whale optimization (LCBWO) structured hybrid framework for predicting heart disease within 5 years using EHR data. Meanwhile, in the hybrid model established multilayer bidirectional LSTM is used for feature selection, LCBWO algorithm for structural improvement and fast convergence, and LSTM for disease prediction. This research used 10 cross‐validations to obtain generalized accuracy and error values. The findings and observations provided here are focused on the knowledge obtained from the EHR report. The results show that the proposed novel hybrid RNN‐LCBWO framework achieves a higher accuracy of 98%, a specificity of 99%, precision of 96%, Mathews correlation coefficient of 91%, F‐measure of 0.9892, an area under the curve value of 98%, and a prediction time of 9.23 seconds. The accurate predictions obtained from the comparative analysis shows the significant performance of our proposed framework.     

A soft computing method to predict sludge volume index based on a recurrent self-organizing neural network

In this paper, a soft computing method, based on a recurrent self-organizing neural network (RSONN) is proposed for predicting the sludge volume index (SVI) in the wastewater treatment process (WWTP). For this soft computing method, a growing and pruning method is developed to tune the structure of RSONN by the sensitivity analysis (SA) of hidden nodes. The redundant hidden nodes will be removed and the new hidden nodes will be inserted when the SA values of hidden nodes meet the criteria. Then, the structure of RSONN is able to be self-organized to maintain the prediction accuracy. Moreover, the convergence of RSONN is discussed in both the self-organizing phase and the phase following the modification of the structure for the soft computing method. Finally, the proposed soft computing method has been tested and compared to other algorithms by applying it to the problem of predicting SVI in WWTP. Experimental results demonstrate its effectiveness of achieving considerably better predicting performance for SVI values.     

A neuro‐based approach to designing a Wilkinson power divider

In this paper, a new neuro‐based approach using a feed‐forward neural network is presented to design a Wilkinson power divider. The proposed power divider is composed of symmetrical modified T‐shaped resonators, which are a replacement for quarter‐wave transmission lines in the conventional structure. The proposed technique reduces the size of the power divider by 45% and suppresses unwanted bands up to the fifth harmonics. To verify the concept, a prototype of the power divider has been fabricated and tested, exhibiting good agreement between the predicted and measured results. The results show that the insertion loss and the isolation at the center frequency are about 3.3 ± 0.1 dB and 23 dB, respectively.