Digital Library

Vol. 17, No. 2, Apr. 2021  Cover

1. Link Prediction Algorithm for Signed Social Networks Based on Local and Global Tightness
   Miao-Miao Liu, Qing-Cui Hu, Jing-Feng Guo, Jing Chen
   Vol. 17, No. 2, pp. 213-226, Apr. 2021  
   https://doi.org/10.3745/JIPS.04.0210
   Details Full Text PubReader
   Keywords: link prediction, Sign Prediction, Signed Social Networks, similarity, Structural Balance Theory, Tightness
   Show / Hide Abstract

   Given that most of the link prediction algorithms for signed social networks can only complete sign prediction, a novel algorithm is proposed aiming to achieve both link prediction and sign prediction in signed networks.Based on the structural balance theory, the local link tightness and global link tightness are defined respectively by using the structural information of paths with the step size of 2 and 3 between the two nodes. Then the total similarity of the node pair can be obtained by combining them. Its absolute value measures the possibility of the two nodes to establish a link, and its sign is the sign prediction result of the predicted link. The effectiveness and correctness of the proposed algorithm are verified on six typical datasets. Comparison and analysis are also carried out with the classical prediction algorithms in signed networks such as CN-Predict, ICN-Predict, and PSNBS (prediction in signed networks based on balance and similarity) using the evaluation indexes like area under the curve (AUC), Precision, improved AUC′, improved Accuracy′, and so on. Results show that the proposed algorithm achieves good performance in both link prediction and sign prediction, and its accuracy is higher than other algorithms. Moreover, it can achieve a good balance between prediction accuracy and computational complexity.


2. RAVIP: Real-Time AI Vision Platform for Heterogeneous Multi-Channel Video Stream
   Jeonghun Lee and Kwang-il Hwang
   Vol. 17, No. 2, pp. 227-241, Apr. 2021  
   https://doi.org/10.3745/JIPS.02.0154
   Details Full Text PubReader
   Keywords: Multi-channel, Multi-Stream, Object Detection, Surveillance Systems, vision
   Show / Hide Abstract

   Object detection techniques based on deep learning such as YOLO have high detection performance and precision in a single channel video stream. In order to expand to multiple channel object detection in real-time, however, high-performance hardware is required. In this paper, we propose a novel back-end server framework, a real-time AI vision platform (RAVIP), which can extend the object detection function from single channel to simultaneous multi-channels, which can work well even in low-end server hardware. RAVIP assembles appropriate component modules from the RODEM (real-time object detection module) Base to create perchannel instances for each channel, enabling efficient parallelization of object detection instances on limited hardware resources through continuous monitoring with respect to resource utilization. Through practical experiments, RAVIP shows that it is possible to optimize CPU, GPU, and memory utilization while performing object detection service in a multi-channel situation. In addition, it has been proven that RAVIP can provide object detection services with 25 FPS for all 16 channels at the same time.


3. Industrial Process Monitoring and Fault Diagnosis Based on Temporal Attention Augmented Deep Network
   Ke Mu, Lin Luo, Qiao Wang, Fushun Mao
   Vol. 17, No. 2, pp. 242-252, Apr. 2021  
   https://doi.org/10.3745/JIPS.04.0211
   Details Full Text PubReader
   Keywords: Deep Learning, Online Fault Classification, Recurrent Neural Networks, Temporal Attention Mechanism
   Show / Hide Abstract

   Following the intuition that the local information in time instances is hardly incorporated into the posterior sequence in long short-term memory (LSTM), this paper proposes an attention augmented mechanism for fault diagnosis of the complex chemical process data. Unlike conventional fault diagnosis and classification methods, an attention mechanism layer architecture is introduced to detect and focus on local temporal information. The augmented deep network results preserve each local instance’s importance and contribution and allow the interpretable feature representation and classification simultaneously. The comprehensive comparative analyses demonstrate that the developed model has a high-quality fault classification rate of 95.49%, on average. The results are comparable to those obtained using various other techniques for the Tennessee Eastman benchmark process.


4. NIST Lightweight Cryptography Standardization Process: Classification of Second Round Candidates, Open Challenges, and Recommendations
   Dennis Agyemanh Nana Gookyi, Guard Kanda, Kwangki Ryoo
   Vol. 17, No. 2, pp. 253-270, Apr. 2021  
   https://doi.org/10.3745/JIPS.03.0156
   Details Full Text PubReader
   Keywords: Authenticated Encryption, CAESAR, IoT, Lightweight cryptography, NIST
   Show / Hide Abstract

   In January 2013, the National Institute of Standards and Technology (NIST) announced the CAESAR (Competition for Authenticated Encryption: Security, Applicability, and Robustness) contest to identify authenticated ciphers that are suitable for a wide range of applications. A total of 57 submissions made it into the first round of the competition out of which 6 were announced as winners in March 2019. In the process of the competition, NIST realized that most of the authenticated ciphers submitted were not suitable for resourceconstrained devices used as end nodes in the Internet-of-Things (IoT) platform. For that matter, the NIST Lightweight Cryptography Standardization Process was set up to identify authenticated encryption and hashing algorithms for IoT devices. The call for submissions was initiated in 2018 and in April 2019, 56 submissions made it into the first round of the competition. In August 2019, 32 out of the 56 submissions were selected for the second round which is due to end in the year 2021. This work surveys the 32 authenticated encryption schemes that made it into the second round of the NIST lightweight cryptography standardization process. The paper presents an easy-to-understand comparative overview of the recommended parameters, primitives, mode of operation, features, security parameter, and hardware/software performance of the 32 candidate algorithms. The paper goes further by discussing the challenges of the Lightweight Cryptography Standardization Process and provides some suitable recommendations.


5. Optimal Two-Section Layouts for the Two-Dimensional Cutting Problem
   Jun Ji, Dun-hua Huang, Fei-fei Xing, Yao-dong Cui
   Vol. 17, No. 2, pp. 271-283, Apr. 2021  
   https://doi.org/10.3745/JIPS.01.0066
   Details Full Text PubReader
   Keywords: Cutting Stock Problem, Layout, Optimization, Two-Dimensional Cutting
   Show / Hide Abstract

   When generating layout schemes, both the material usage and practicality of the cutting process should be considered. This paper presents a two-section algorithm for generating guillotine-cutting schemes of rectangular blanks. It simplifies the cutting process by allowing only one size of blanks to appear in any rectangular block. The algorithm uses an implicit enumeration and a linear programming optimal cutting scheme to maximize the material usage. The algorithm was tested on some benchmark problems in the literature, and compared with the three types of layout scheme algorithm. The experimental results show that the algorithm is effective both in computation time and in material usage.


6. Reference Architecture and Operation Model for PPP (Public-Private-Partnership) Cloud
   Youngkon Lee, Ukhyun Lee
   Vol. 17, No. 2, pp. 284-296, Apr. 2021  
   https://doi.org/10.3745/JIPS.04.0212
   Details Full Text PubReader
   Keywords: Cloud Operational Model, Cloud Reference Model, PPP Cloud
   Show / Hide Abstract

   The cloud has already become the core infrastructure of information systems, and government institutions are rapidly migrating information systems to the cloud. Government institutions in several countries use private clouds in their closed networks. However, because of the advantages of public clouds over private clouds, the demand for public clouds is increasing, and government institutions are expected to gradually switch to public clouds. When all data from government institutions are managed in the public cloud, the biggest concern for government institutions is the leakage of confidential data. The public-private-partnership (PPP) cloud provides a solution to this problem. PPP cloud is a form participation in a public cloud infrastructure and the building of a closed network data center. The PPP cloud prevents confidential data leakage and leverages the benefits of the public cloud to build a cloud quickly and easily maintain the cloud. In this paper, based on the case of the PPP cloud applied to the Korean government, the concept, architecture, operation model, and contract method of the PPP cloud are presented.


7. An Efficient Load Balancing Scheme for Gaming Server Using Proximal Policy Optimization Algorithm
   Hye-Young Kim
   Vol. 17, No. 2, pp. 297-305, Apr. 2021  
   https://doi.org/10.3745/JIPS.03.0158
   Details Full Text PubReader
   Keywords: Dynamic allocation, greedy algorithm, load balancing, proximal policy optimization, Reinforcement Learning
   Show / Hide Abstract

   Large amount of data is being generated in gaming servers due to the increase in the number of users and the variety of game services being provided. In particular, load balancing schemes for gaming servers are crucial consideration. The existing literature proposes algorithms that distribute loads in servers by mostly concentrating on load balancing and cooperative offloading. However, many proposed schemes impose heavy restrictions and assumptions, and such a limited service classification method is not enough to satisfy the wide range of service requirements. We propose a load balancing agent that combines the dynamic allocation programming method, a type of greedy algorithm, and proximal policy optimization, a reinforcement learning. Also, we compare performances of our proposed scheme and those of a scheme from previous literature, ProGreGA, by running a simulation.


8. Power Allocation Method of Downlink Non-orthogonal Multiple Access System Based on α Fair Utility Function
   Jianpo Li, Qiwei Wang
   Vol. 17, No. 2, pp. 306-317, Apr. 2021  
   https://doi.org/10.3745/JIPS.03.0157
   Details Full Text PubReader
   Keywords: Ergodic Sum Rate, NOMA, power allocation, α Fair Utility Function
   Show / Hide Abstract

   The unbalance between system ergodic sum rate and high fairness is one of the key issues affecting the performance of non-orthogonal multiple access (NOMA) system. To solve the problem, this paper proposes a power allocation algorithm to realize the ergodic sum rate maximization of NOMA system. The scheme is mainly achieved by the construction algorithm of fair model based on α fair utility function and the optimal solution algorithm based on the interior point method of penalty function. Aiming at the construction of fair model, the fair target is added to the traditional power allocation model to set the reasonable target function. Simultaneously, the problem of ergodic sum rate and fairness in power allocation is weighed by adjusting the value of α. Aiming at the optimal solution algorithm, the interior point method of penalty function is used to transform the fair objective function with unequal constraints into the unconstrained problem in the feasible domain. Then the optimal solution of the original constrained optimization problem is gradually approximated within the feasible domain. The simulation results show that, compared with NOMA and time division multiple address (TDMA) schemes, the proposed method has larger ergodic sum rate and lower Fairness Index (FI) values.


9. Towards a Redundant Response Avoidance for Intelligent Chatbot
   Hyuck-Moo Gwon, Yeong-Seok Seo
   Vol. 17, No. 2, pp. 318-333, Apr. 2021  
   https://doi.org/10.3745/JIPS.04.0213
   Details Full Text PubReader
   Keywords: Chatbot, Human-Computer Interaction, Interactive System, Redundancy Avoidance, Telegram
   Show / Hide Abstract

   Smartphones are one of the most widely used mobile devices allowing users to communicate with each other. With the development of mobile apps, many companies now provide various services for their customers by studying interactive systems in the form of mobile messengers for business marketing and commercial promotion. Such interactive systems are called “chatbots.” In this paper, we propose a method of avoiding the redundant responses of chatbots, according to the utterances entered by the user. In addition, the redundant patterns of chatbot responses are classified into three categories for the first time. In order to verify the proposed method, a chatbot is implemented using Telegram, an open source messenger. By comparing the proposed method with an existent method for each pattern, it is confirmed that the proposed method significantly improves the redundancy avoidance rate. Furthermore, response performance and variation analysis of the proposed method are investigated in our experiment


10. A Special Section on Deep & Advanced Machine Learning Approaches for Human Behavior Analysis
    Yizhang Jiang, Kim-Kwang Raymond Choo, Hoon Ko
    Vol. 17, No. 2, pp. 334-336, Apr. 2021  
    https://doi.org/10.3745/JIPS.01.0074
    Details Full Text PubReader
    Keywords:
    Show / Hide Abstract


11. Video Expression Recognition Method Based on Spatiotemporal Recurrent Neural Network and Feature Fusion
    Xuan Zhou
    Vol. 17, No. 2, pp. 337-351, Apr. 2021  
    https://doi.org/10.3745/JIPS.01.0067
    Details Full Text PubReader
    Keywords: Double Layer Cascade Structure, facial expression recognition, feature fusion, Image Detection, Spatiotemporal Recursive Neural Network
    Show / Hide Abstract

    Automatically recognizing facial expressions in video sequences is a challenging task because there is little direct correlation between facial features and subjective emotions in video. To overcome the problem, a video facial expression recognition method using spatiotemporal recurrent neural network and feature fusion is proposed. Firstly, the video is preprocessed. Then, the double-layer cascade structure is used to detect a face in a video image. In addition, two deep convolutional neural networks are used to extract the time-domain and airspace facial features in the video. The spatial convolutional neural network is used to extract the spatial information features from each frame of the static expression images in the video. The temporal convolutional neural network is used to extract the dynamic information features from the optical flow information from multiple frames of expression images in the video. A multiplication fusion is performed with the spatiotemporal features learned by the two deep convolutional neural networks. Finally, the fused features are input to the support vector machine to realize the facial expression classification task. The experimental results on cNTERFACE, RML, and AFEW6.0 datasets show that the recognition rates obtained by the proposed method are as high as 88.67%, 70.32%, and 63.84%, respectively. Comparative experiments show that the proposed method obtains higher recognition accuracy than other recently reported methods.


12. POI Recommendation Method Based on Multi-Source Information Fusion Using Deep Learning in Location-Based Social Networks
    Liqiang Sun
    Vol. 17, No. 2, pp. 352-368, Apr. 2021  
    https://doi.org/10.3745/JIPS.01.0068
    Details Full Text PubReader
    Keywords: Convolutional Neural Network, Emotional Information, Geographical Information, Latent Factor Modelling, Location-Based Social Network, objective function
    Show / Hide Abstract

    Sign-in point of interest (POI) are extremely sparse in location-based social networks, hindering recommendation systems from capturing users’ deep-level preferences. To solve this problem, we propose a content-aware POI recommendation algorithm based on a convolutional neural network. First, using convolutional neural networks to process comment text information, we model location POI and user latent factors. Subsequently, the objective function is constructed by fusing users’ geographical information and obtaining the emotional category information. In addition, the objective function comprises matrix decomposition and maximisation of the probability objective function. Finally, we solve the objective function efficiently. The prediction rate and F1 value on the Instagram-NewYork dataset are 78.32% and 76.37%, respectively, and those on the Instagram-Chicago dataset are 85.16% and 83.29%, respectively. Comparative experiments show that the proposed method can obtain a higher precision rate than several other newer recommended methods.


13. Personalized Product Recommendation Method for Analyzing User Behavior Using DeepFM
    Jianqiang Xu, Zhujiao Hu, Junzhong Zou
    Vol. 17, No. 2, pp. 369-384, Apr. 2021  
    https://doi.org/10.3745/JIPS.01.0069
    Details Full Text PubReader
    Keywords: DeepFM, Higher-Order Feature, Hit Rate Prediction, K-Means Similarity Clustering, Low-Order Features, Personalized Product Recommendation
    Show / Hide Abstract

    In a personalized product recommendation system, when the amount of log data is large or sparse, the accuracy of model recommendation will be greatly affected. To solve this problem, a personalized product recommendation method using deep factorization machine (DeepFM) to analyze user behavior is proposed. Firstly, the K-means clustering algorithm is used to cluster the original log data from the perspective of similarity to reduce the data dimension. Then, through the DeepFM parameter sharing strategy, the relationship between low- and high-order feature combinations is learned from log data, and the click rate prediction model is constructed. Finally, based on the predicted click-through rate, products are recommended to users in sequence and fed back. The area under the curve (AUC) and Logloss of the proposed method are 0.8834 and 0.0253, respectively, on the Criteo dataset, and 0.7836 and 0.0348 on the KDD2012 Cup dataset, respectively. Compared with other newer recommendation methods, the proposed method can achieve better recommendation effect.


14. Residual Learning Based CNN for Gesture Recognition in Robot Interaction
    Hua Han
    Vol. 17, No. 2, pp. 385-398, Apr. 2021  
    https://doi.org/10.3745/JIPS.01.0072
    Details Full Text PubReader
    Keywords: Convolutional Neural Network, Feature Redundancy, Full Connection Layer, Gesture Recognition, Human- Computer Interaction, residual learning
    Show / Hide Abstract

    The complexity of deep learning models affects the real-time performance of gesture recognition, thereby limiting the application of gesture recognition algorithms in actual scenarios. Hence, a residual learning neural network based on a deep convolutional neural network is proposed. First, small convolution kernels are used to extract the local details of gesture images. Subsequently, a shallow residual structure is built to share weights, thereby avoiding gradient disappearance or gradient explosion as the network layer deepens; consequently, the difficulty of model optimisation is simplified. Additional convolutional neural networks are used to accelerate the refinement of deep abstract features based on the spatial importance of the gesture feature distribution. Finally, a fully connected cascade softmax classifier is used to complete the gesture recognition. Compared with the dense connection multiplexing feature information network, the proposed algorithm is optimised in feature multiplexing to avoid performance fluctuations caused by feature redundancy. Experimental results from the ISOGD gesture dataset and Gesture dataset prove that the proposed algorithm affords a fast convergence speed and high accuracy.


15. A Facial Expression Recognition Method Using Two-Stream Convolutional Networks in Natural Scenes
    Lixin Zhao
    Vol. 17, No. 2, pp. 399-410, Apr. 2021  
    https://doi.org/10.3745/JIPS.01.0070
    Details Full Text PubReader
    Keywords: Attentional Mechanism, Confrontational Learning, Double Flow Convolutional Neural Network, Image Preprocessing, Natural Scene Expression Recognition
    Show / Hide Abstract

    Aiming at the problem that complex external variables in natural scenes have a greater impact on facial expression recognition results, a facial expression recognition method based on two-stream convolutional neural network is proposed. The model introduces exponentially enhanced shared input weights before each level of convolution input, and uses soft attention mechanism modules on the space-time features of the combination of static and dynamic streams. This enables the network to autonomously find areas that are more relevant to the expression category and pay more attention to these areas. Through these means, the information of irrelevant interference areas is suppressed. In order to solve the problem of poor local robustness caused by lighting and expression changes, this paper also performs lighting preprocessing with the lighting preprocessing chain algorithm to eliminate most of the lighting effects. Experimental results on AFEW6.0 and Multi-PIE datasets show that the recognition rates of this method are 95.05% and 61.40%, respectively, which are better than other comparison methods.


16. Vehicle Image Recognition Using Deep Convolution Neural Network and Compressed Dictionary Learning
    Yanyan Zhou
    Vol. 17, No. 2, pp. 411-425, Apr. 2021  
    https://doi.org/10.3745/JIPS.01.0073
    Details Full Text PubReader
    Keywords: Adaptive Fusion, Compressed Dictionary Learning, Deep Convolutional Neural Network, High-Dimensional Features, Vehicle Type Recognition
    Show / Hide Abstract

    In this paper, a vehicle recognition algorithm based on deep convolutional neural network and compression dictionary is proposed. Firstly, the network structure of fine vehicle recognition based on convolutional neural network is introduced. Then, a vehicle recognition system based on multi-scale pyramid convolutional neural network is constructed. The contribution of different networks to the recognition results is adjusted by the adaptive fusion method that adjusts the network according to the recognition accuracy of a single network. The proportion of output in the network output of the entire multiscale network. Then, the compressed dictionary learning and the data dimension reduction are carried out using the effective block structure method combined with very sparse random projection matrix, which solves the computational complexity caused by highdimensional features and shortens the dictionary learning time. Finally, the sparse representation classification method is used to realize vehicle type recognition. The experimental results show that the detection effect of the proposed algorithm is stable in sunny, cloudy and rainy weather, and it has strong adaptability to typical application scenarios such as occlusion and blurring, with an average recognition rate of more than 95%.


17. Personalized Web Service Recommendation Method Based on Hybrid Social Network and Multi-Objective Immune Optimization
    Huashan Cao
    Vol. 17, No. 2, pp. 426-439, Apr. 2021  
    https://doi.org/10.3745/JIPS.01.0071
    Details Full Text PubReader
    Keywords: Cold Boot, Hybrid Social Networks, Personalized Recommendation, Multi-Objective Immune Optimization, Service Providers, Web Service Recommendation
    Show / Hide Abstract

    To alleviate the cold-start problem and data sparsity in web service recommendation and meet the personalized needs of users, this paper proposes a personalized web service recommendation method based on a hybrid social network and multi-objective immune optimization. The network adds the element of the service provider, which can provide more real information and help alleviate the cold-start problem. Then, according to the proposed service recommendation framework, multi-objective immune optimization is used to fuse multiple attributes and provide personalized web services for users without adjusting any weight coefficients. Experiments were conducted on real data sets, and the results show that the proposed method has high accuracy and a low recall rate, which is helpful to improving personalized recommendation.