Computing and Informatics (E-Journal - Institute of Informatics, SAS, Bratislava)
Not a member yet
1506 research outputs found
Sort by
Multi-Modal Emotion Recognition Using Situation-Based Video Context Emotion Dataset
No full textCurrent multi-modal emotion recognition techniques primarily use modalities such as expression, speech, text, and gesture. Existing methods only capture emotion from the current moment in a picture or video, neglecting the influence of time and past experiences on human emotion. Expanding the temporal scope can provide more clues for emotion recognition. To address this, we constructed the Situation-Based Video Context Emotion Datasets (SVCEmotion) dataset in video form. Experiments show that both VGGish and BERTbase achieve good results on SVCEmotion. Comparison with other audio emotion recognition methods proves that VGGish is more suitable for audio emotion feature extraction on the dataset constructed in this paper. Comparison experiments with textual descriptions demonstrate that the contextual descriptions introduced in the SVCEmotion dataset for the emotion recognition task under wide time range can provide clues for emotion recognition, and that the combination with factual descriptions can substantially improve the emotion recognition effect
Data-Driven Bayesian Network for Risk Analysis of Telecom Fraud
Full text linkGiven the widespread occurrence of global telecom fraud, the development of proactive measures for crime prevention and control has become increasingly crucial. This study introduces a data-driven Bayesian Network (BN) model, which incorporates D-S evidence theory to integrate prior knowledge for fraud risk analysis. Through the examination of real-world case data, the study identifies key risk-influencing factors (RIFs) and uncovers causal relationships by comparatively evaluating three structure learning algorithms: Peter-Clark (PC), Bayesian Search (BS), and Greedy Thick Thinning (GTT). A robust Directed Acyclic Graph (DAG) is then constructed, and the Expectation-Maximization (EM) algorithm is employed to estimate conditional probability distributions. The proposed model effectively captures the causal relationships and nonlinear complexities among RIFs. To validate the model's applicability, scenario reasoning and sensitivity analysis are conducted, confirming its effectiveness in prioritizing RIFs and supporting informed decision-making. This research presents a novel and practical framework for public security agencies to develop proactive strategies for telecom fraud prevention and control
Modeling and Analyzing Hormonal Effects of Depression Based on Petri Nets and Machine Learning
Full text linkDepression has become a common mental illness, and the number of patients has shown a noticeable rising trend. However, the exploration of the connection between hormone levels and physical state changes in depression patients is still open. Hormone levels are complex and play a key role in regulating multiple body systems and functions, directly or indirectly influencing overall health and physical state. This work utilizes Petri nets to establish a corresponding model for the transition of hormone levels and states in depression, focusing on the association between different hormone levels and states in depressive patients. At the same time, machine learning methods offer a new approach to predicting the reachability of depression patients' states. This work enables healthcare professionals to quickly assess patients' emotional changes and their impact on outcomes, improving resource allocation
HiBiGNN: Hierarchical Bilateral Graph Neural Network for fMRI Analysis
No full textGraph Neural Networks (GNNs) have shown great promise in functional Magnetic Resonance Imaging (fMRI) analysis due to their ability to capture complex interactions between brain regions. However, existing models often overlook the brain's physiological structure and fail to leverage hierarchical information from brain atlases. In this paper, we propose Hierarchical Bilateral Graph Neural Network (HiBiGNN), a generic architecture that integrates hierarchical information from brain atlases and incorporates the bilateral structure of the brain, with the ability to be instantiated with various existing GNNs as its foundation. HiBiGNN processes a special heterogeneous graph structure, called the Hierarchical Bilateral Graph (HiBiG), which combines multi-level brain graphs derived from functional regions defined by multi-level brain atlases and divides each brain graph into left and right subgraphs, thereby modeling multiple types of nodes and relations. During feature extraction, HiBiGNN performs deep fusion of features from different types of nodes using a unique convolution operation (HiBiG-Conv) and generates graph-level representations via a specialized readout operation (HiBiG-Readout) for graph classification tasks. To assess the effectiveness of HiBiGNN, we conducted extensive experiments on a graph classification task using an fMRI dataset we collected from a response inhibition task, testing multiple HiBiGNN instances with different base GNN models. The results show that our HiBiGNN instances outperforms several generic GNN models as well as those specifically designed for fMRI analysis, demonstrating the significant potential of HiBiGNN for future applications
Block-Jacobi SVD Algorithms: A Review
No full textWe discuss some progress in the design and implementation of block-Jacobi SVD algorithms in last 25 years. Two ideas were crucial for enhancing the efficiency of both two- and one-sided, serial or parallel, block-Jacobi methods: the so-called dynamic ordering of subproblems solved in each iteration step, and a suitable preconditioning of an original matrix. These two ideas led to a substantial decrease of (serial or parallel) iteration steps needed for the convergence. Consequently, especially the one-sided block-Jacobi SVD algorithm became competitive in speed with some SVD algorithms based on the matrix bi-diagonalization. We also discuss new theoretical results w.r.t. the asymptotic quadratic convergence of block-Jacobi SVD algorithms regardless to the distribution of singular values of an original matrix
Adaptive Non-Overlapping Community Detection Based on Gravitational Field Stability in Social Networks
Full text linkCommunity structure is a common feature of social networks and many community discovery algorithms have emerged through the study of this feature. The gravitational field model is an effective method to realize community division. However, the current gravitational field model lacks a comprehensive consideration of field properties such as the internal stability of the gravitational field. Therefore, in this paper, we define and quantify the attributes of the gravitational field by taking advantage of the field's strength in describing the joint action of groups. Then, we propose a social network gravitational field community detection model (GF-CDM). GF-CDM selects the field kernel node based on a random walk and then presents an adaptive expansion function of fusion field stability to divide the observable network into overlapping and non-overlapping clusters. The model was evaluated on four real network datasets and five artificial network datasets of different sizes. Experimental results show that our proposed model outperforms the other four benchmark algorithms in modularity, ARI index, and field average stability, which can improve the quality of cluster division
Personalized Federated Learning Based on Hypernetworks and Attention Mechanism Ensembles for Internet of Things
No full textAs the demand for data privacy protection continues to grow and the concept of collaborative modeling gains traction, federated learning has emerged as a pivotal distributed learning paradigm in the Internet of Things (IoT) domain. However, the client data held by different institutions often varies significantly in sources and characteristics, which can hinder the efficiency of federated learning model training and increase the risk of personal privacy breaches. To address the challenges of model accuracy degradation and privacy exposure when federated learning is applied to multi-source heterogeneous data, we propose a personalized federated learning strategy that integrates hypernetworks with attention mechanisms. This strategy involves transforming labeled data at the source to protect personal privacy while employing hypernetworks and Transformer-based mechanisms to focus on the personalized information of clients from various institutions. Our proposed approach supports handling heterogeneous data, thereby better meeting the personalized needs of different institutions. Experimental results demonstrate that this framework not only effectively safeguards data privacy but also significantly enhances the performance and generalization capability of federated learning on heterogeneous data. This research offers a novel perspective for developing more adaptable personalized federated learning models, facilitating cross-institutional collaborative research, and providing an innovative model training solution for various IoT devices, balancing the dual requirements of data privacy protection and multi-institutional data sharing
Multi-Objective Optimization for Multi-Modal Route Planning Integrating Shared Taxi and Bus
Full text linkMulti-modal transportation, emerging as a sustainable travel option, has shown immense promise in reducing passengers' travel expenses and vehicles' energy consumption. To further promote green travel, this work studies a multi-modal route planning problem, focusing on the integration of shared taxis and buses. The objective is to devise an innovative route planning approach for shared taxis, enabling passengers a seamless transition between the two modes and arrive at their destinations within designated timeframes. It designs a new pricing rule and establishes a multi-objective optimization that takes into account both the interests of passengers and shared taxi operators. The objectives are to minimize the aggregate cost incurred by all passengers and the overall travel distance traversed by shared taxis, and maximizes the revenue earned per kilometer by shared taxi operators. A novel nondominated linear sorting genetic algorithm (NLSGA) is introduced to tackle the problem. This algorithm incorporates innovative evolution and selection strategies to preserve solution diversity and enhance convergence speed. NLSGA demonstrates superior performance compared to several widely used multi-objective optimization algorithms, including NSGA-II, MOPSO, and MOGWO. Experimental results reveal that the proposed algorithm effectively reduces passengers' cost and shared taxis' travel distance while simultaneously maximizing revenue per kilometer for shared taxi operators
Semantic Enhancement and Heterogeneous Correlation Guided Web Service Clustering
Full text linkService description suffers from short texts and contains few repeated words, which brings challenges to generate high-quality service function vector (SFV) in service clustering. Some works introduce service association to improve service clustering quality. However, they simply introduce single associations, such as tag associations or collaboration association. Single service association can only improve the clustering quality from one perspective of positive or negative categorical relevance. In this study, we propose semantic enhancement and heterogeneous correlation guided Web Service Clustering. A high-performance contrastive learning framework is employed to generate SFVs. Meanwhile, we propose a method for the semantic enhancement of SFVs by obtaining twin service descriptions through verb substitution. A heterogeneous association is established based on tag association and collaboration association. It quantitatively enhances the clustering quality from both positive and negative categorical relevance. Experiments show that the proposed method outperforms popular semantic enhancement ways in generating high-quality SFVs. The heterogeneous association can significantly improve service clustering quality compared to single tag association or collaborative association. The clustering quality obtained by our method is improved by 13.7 %, 9 %, 6.8 %, 6.1 %, and 5.5 % on average over the state-of-the-art service clustering methods in terms of DBI, SC, AMI, NMI, and Purity
Personalized Learning Path Recommendation Based on Learner Profile and Knowledge Graph
Full text linkE-learning is increasingly popular because it allows learners to freely choose their class times and locations. However, traditional E-learning platforms face issues of information overload and fragmented resources. The proposition of the concept of personalized learning has effectively alleviated these problems. However, current personalized learning recommendation methods fail to comprehensively and systematically address learners' needs. To solve this issue, this paper proposes a learning path recommendation method based on learner profiles. First, by collecting learners' personal information, learning history, and behavior data, a learner profile is established considering multiple aspects. Then, generating a path evaluation function for learners from the profile. Using the Ant Colony Optimization algorithm, the most suitable personalized learning path for the learner's needs is searched within the knowledge graph. Experimental results demonstrate that the personalized learning path recommendations generated by our algorithm meet expectations and achieve the best overall performance in comparative experiments