Computing and Informatics (E-Journal - Institute of Informatics, SAS, Bratislava)
Not a member yet
1506 research outputs found
Sort by
Bonding Defect Detection Based on Improved Single Shot MultiBox Detector
Full text linkTo solve the problem of time-consuming and low efficiency in manual defect detection, this paper proposes a bonding defect detection algorithm based on improved Single Shot MultiBox Detector (SSD). DenseNet is used to replace VGG of the SSD algorithm to improve the detection effect of bonding defect. A novel feature fusion network is designed, in which dilated convolution is used to reduce the size of the low-level feature map, and it is fused with the high-level feature map, and then the Convolutional Block Attention Module (CBAM) attention mechanism is used to increase the ability to extract the features. Focal loss is used to control the ratio of positive and negative samples for training and suppress easily separable samples, so that the samples involved in training have better distribution and the model has better detection performance. Then, the defect data set is constructed and a comparison experiment is carried out. The results show that the mAP, Precision, and Recall of the improved SSD network are increased to 75.9 %, 77.3 %, and 75.6 %, respectively, which can better identify bonding defect
DifBFSR: Blind Face Super-Resolution via Conditional Diffusion Contraction
Full text linkBlind Face Super-Resolution (BFSR) has recently gained widespread attention, which aims to super-resolve Low-Resolution (LR) face images with complex unknown degradation to High-Resolution (HR) face images. However, existing BFSR methods suffer from two major limitations. First, most of them are trained on synthetic degradation data pairs with pre-defined degradation models, which leads to poor performance due to the degradation mismatch between other unknown complex degradations in real-world scenarios. Second, some methods rely on hand-crafted face priors as constraints, such as facial landmarks and parsing maps, which require additional callouts and laborious hyperparameter tuning for real cases. To tackle these issues, we propose a simple and effective self-supervised cooperative learning framework via a conditional diffusion contraction method for BFSR, dubbed DifBFSR, which establishes the posterior distribution of HR images from degraded LR images with unknown degradation via a powerful diffusion model without expensive supervised training or additional constraint design. Specifically, we first transform the degraded LR face image to an intermediate HR face prediction with degradation-invariant by a simple Super-Resolution module (SRM), which only relies on self-supervised optimization. To enhance the face prediction, we propose a Contraction Filter Module (CFM) to gradually contract the restoration error by adaptive dynamic filtering, which efficiently leverages rich nature face prior encapsulated in the pre-trained diffusion model through conditional posterior sampling. Finally, by combining the SRM, CFM, and diffusion model in a self-supervised cooperative learning framework, DifBFSR can robustly handle unknown complex degradations, which favorably avoids the cumbersome training and parameter tuning. Extensive qualitative and quantitative experiments on complex degraded synthetic and real-world datasets show that our method outperforms state-of-the-art BFSR methods
FedDRL: Trustworthy Federated Learning Model Fusion Method Based on Staged Reinforcement Learning
Full text linkFederated learning facilitates collaborative data analysis among multiple participants while preserving user privacy. However, conventional federated learning approaches, typically employing weighted average techniques for model fusion, confront two significant challenges: 1. The inclusion of malicious models in the fusion process can drastically undermine the accuracy of the aggregated global model. 2. Due to the heterogeneity problem of devices and data, the number of client samples does not determine the weight value of the model. To solve those challenges, we propose a trustworthy model fusion method based on reinforcement learning (FedDRL), which includes two stages. In the first stage, we propose a reliable client selection mechanism to exclude malicious models from the fusion process. In the second stage, we propose an adaptive model fusion method that dynamically assigns weights based on model quality to aggregate the best global models. Finally, we validate our approach against five distinct model fusion scenarios, demonstrating that our algorithm significantly enhanced reliability without compromising accuracy
Identification of KLF9 and FOSL2 as Endoplasmic Reticulum Stress Signature Genes in Osteoarthritis with Multiple Machine Learning Approaches
Full text linkObjective: This study aims to screen osteoarthritis (OA) endoplasmic reticulum (ER) stress signature genes using a machine learning approach to provide new insights and methods for OA treatment. Methods: We obtained GSE55235 and GSE98918 datasets from the gene expression omnibus (GEO) database and identified ER stress-related genes from the GeneCard database. We used R software to perform data batch correction, extract OA endoplasmic reticulum stress-related genes, and conduct differential analysis. We performed functional Gene Ontology (GO) analysis, Kyoto Encyclopedia of Genes and Genomes (KEGG) signaling pathway analysis, and gene set enrichment analysis (GSEA) on differentially expressed genes (DEGs). Additionally, we used machine learning algorithms, including Least Absolute Shrinkage and Selection Operator (LASSO) regression, SVM-RFE, and weighted gene co-expression network analysis (WGCNA), to screen OA endoplasmic reticulum stress signature genes. Human chondrocytes were selected for OA model establishment, cells without any treatment were served as the control. Results: We obtained 236 DEGs related to OA ER stress. GO and KEGG enrichment analysis showed that these genes were mainly involved in the positive regulation of leukocyte activation, collagen-containing extracellular matrix, phagosome, and other biological functions or signaling pathways. GSEA-GO analysis revealed that ER stress genes were significantly enriched in the negative regulation in metabolic processes of nucleobase-containing compounds (NES = -2.50, P < 0.001), while OA ER stress genes were significantly enriched in the processing and presentation of peptide antigens (NES = 2.40, P < 0.001). Using WGCNA analysis, LASSO regression analysis, and SVM-RFE analysis of intersection, we identified KLF9 and FOSL2 as potential OA endoplasmic reticulum stress signature genes, which were found to be more accurate as OA signature genes after validation. KLF9 expression in OA group was higher than that in control group, while FOSL2 expression was lower (P < 0.05). Conclusion: Machine learning and co-expression network analysis can effectively identify the genes and potential factors characteristic of ER stress in OA, which can help elucidate its pathogenesis and provide a new direction for better clinical treatment
Evaluation and Application Algorithm of Artificial Intelligence Unmanned Vehicle Control Device Based on IoT Intelligent Transportation
Full text linkWith the rapid development of technology, unmanned vehicles have become a hot research topic in the field of intelligent transportation. Unmanned vehicles have many advantages, such as improving traffic efficiency, reducing traffic accidents, and reducing energy consumption. However, the controllability and safety of unmanned vehicles have always been a key issue in research. The Internet of Things can achieve information exchange and data sharing between vehicles, transportation facilities, traffic management centers, etc., providing real-time traffic and road condition information, and providing accurate data support for intelligent decision-making and path planning of unmanned vehicles. Therefore, the artificial intelligence unmanned vehicle control device based on IoT intelligent transportation has become an important research direction. This paper used deep reinforcement learning as the decision-making control algorithm, and designed a set of unmanned vehicle control system based on the DDPG (Deep Deterministic Policy Grad) algorithm, supplemented by meta DDPG algorithm, which is the knowledge of meta learning. Through the test of the simulation platform, it was concluded that the control system has good generalization. This study combined the Internet of Things and artificial intelligence algorithms, which has certain guiding significance for the development and application of unmanned vehicles in the future
Adaptive Mathematical Morphology with Fuzzy Structuring Element
Full text linkAs a well-known nonlinear tool, mathematical morphology (MM) is still active in image processing. Benefiting from the fixed structuring element (SE), traditional MM (TMM) gets solid theoretical foundation. However, due to the inherent diversity of pixels in an image, the rigid SE paradigm is not always practical. As a result, the development of morphology with adaptive SE, known as adaptive MM (AMM), has been a significant challenge. In this work, we present a novel approach for designing adaptive SE using the \boldsymbolα-cut of a fuzzy set. By implementing dilation and erosion operations serially, we obtain an AMM (named SAMM) that is both adaptive to image content and robust to noise. Additionally, the operators in SAMM inherit important properties from TMM as much as possible. We provide theoretical proofs and simulated results to support our conclusion. Preliminary experiments on edge detection and noise reduction confirm the effectiveness of our SAMM both quantitatively and perceptually. In the denoising experiments, SAMM achieves the best performance in the nine algorithms involved, with its PSNR value surpassing the second-ranked approach by more than 0.6 dB overall. Additionally, its SSIM quantification metric also ranks prominently among the top performers
Multi-Stream Convolutional Neural Network with Frequency Selection for Robust Speaker Verification
Full text linkSpeaker verification aims to verify whether an input speech corresponds to the claimed speaker, and conventionally, this kind of system is deployed based on single-stream scenario, wherein the feature extractor operates in full frequency range. In this paper, we hypothesize that machine can learn enough knowledge to do classification task when listening to partial frequency range instead of full frequency range, which is so called frequency selection technique, and further propose a novel framework of multi-stream Convolutional Neural Network (CNN) with this technique for speaker verification tasks. The proposed framework accommodates diverse temporal embeddings generated from multiple streams to enhance the robustness of acoustic modeling. For the diversity of temporal embeddings, we consider feature augmentation with frequency selection, which is to manually segment the full-band of frequency into several sub-bands, and the feature extractor of each stream can select which sub-bands to use as target frequency domain. Different from conventional single-stream solution wherein each utterance would only be processed for one time, in this framework, there are multiple streams processing it in parallel. The input utterance for each stream is pre-processed by a frequency selector within specified frequency range, and post-processed by mean normalization. The normalized temporal embeddings of each stream will flow into a pooling layer to generate fused embeddings. We conduct extensive experiments on VoxCeleb dataset, and the experimental results demonstrate that multi-stream CNN significantly outperforms single-stream baseline with 20.53 % of relative improvement in minimum Decision Cost Function (minDCF) and 15.28 % of relative improvement in Equal Error Rate (EER)
Multi-Agent Dynamic Leader-Follower Path Planning Applied to the Multi-Pursuer Multi-Evader Game
Full text linkMulti-agent collaborative path planning focuses on how the agents have to coordinate their displacements in the environment to achieve different targets or to cover a specific zone in a minimum of time. Reinforcement learning is often used to control the agents' trajectories in the case of static or dynamic targets. In this paper, we propose a multi-agent collaborative path planning based on reinforcement learning and leader-follower principles. The main objectives of this work are the development of an applicable motion planning in a partially observable environment, and also, to improve the agents' cooperation level during the tasks' execution via the creation of a dynamic hierarchy in the pursuit groups. This dynamic hierarchy is reflected by the possibility of reattributing the roles of Leaders and Followers at each iteration in the case of mobile agents to decrease the task's execution time. The proposed approach is applied to the Multi-Pursuer Multi-Evader game in comparison with recently proposed path planning algorithms dealing with the same problem. The simulation results reflect how this approach improves the pursuit capturing time and the payoff acquisition during the pursuit
Corporate Fraud Detection Based on Improved BP Neural Network
Full text linkCorporate fraud risk detection is a branch of fraud. It may exist in various industries and cause economic problems. Effective identification of corporate fraud can protect the safety of funds for investors in some sense. This paper proposes a classifier model of a fractional-order immune BP neural network based on the self-attention mechanism to improve efficiency. The improved artificial immune algorithm with dynamic region contraction strategy is used to optimize the initialization process of the BP neural network. Furthermore, it combines the self-attention mechanism to design the input layer. Finally, Caputo fractional non-causal calculus is used to optimize the parameter updating process in BP neural network. The experiment results indicate that our model has fast convergence rate and powerful capacity of detection, and performs efficiently in detecting fraud behaviors
MGRF: Multi-Graph Recommendation Framework with Heterogeneous and Homogeneous Graph Iterative Fusion
Full text linkWith the development of deep learning, deep neural methods have been introduced to boost the performance of Collaborative Filtering (CF) models. However, most of the models rely solely on the user-item heterogeneous graph and only implicitly capture homogenous information, which limits their performance improvement. Although some state-of-the-art methods try to utilize additional graphs to make up, they either merely aggregate the information of multiple graphs in the step of initial embedding or only merge different multi-graph information in the step of final embedding. Such one-time multi-graph integration leads to the loss of interactive and topological information in the intermediate process of propagation. This paper proposes a novel Multi-Graph iterative fusion Recommendation Framework (MGRF) for CF recommendation. The core components are dual information crossing interaction and multi-graph fusing propagation. The former enables repeated feature crossing between heterogeneous nodes throughout the whole embedding process. The latter repeatedly integrates homogeneous nodes as well as their topological relationships based on the constructed user-user and item-item graphs. Thus, MGRF can improve the embedding quality by iteratively fusing user-item heterogeneous graph, user-user and item-item homogeneous graphs. Extensive experiments on three public benchmarks demonstrate the effectiveness of MGRF, which outperforms state-of-the-art baselines in terms of Recall and NDCG.