Map Network Research Articles

Multi-layer graph attention neural networks for accurate drug-target interaction mapping

In the crucial process of drug discovery and repurposing, precise prediction of drug-target interactions (DTIs) is paramount. This study introduces a novel DTI prediction approach—Multi-Layer Graph Attention Neural Network (MLGANN), through a groundbreaking computational framework that effectively harnesses multi-source information to enhance prediction accuracy. MLGANN not only strides forward in constructing a multi-layer DTI network by capturing both direct interactions between drugs and targets as well as their multi-level information but also amalgamates Graph Convolutional Networks (GCN) with a self-attention mechanism to comprehensively integrate diverse data sources. This method exhibited significant performance surpassing existing approaches in comparative experiments, underscoring its immense potential in elevating the efficiency and accuracy of DTI predictions. More importantly, this study accentuates the significance of considering multi-source data information and network heterogeneity in the drug discovery process, offering new perspectives and tools for future pharmaceutical research.

Loneliness and Social Isolation: qualitative study among youth in vulnerable situations

Abstract Background Loneliness and social isolation are increasing among young people, particularly those in vulnerable NEET (Not in Education, Employment or Training) situations, potentially leading to poor health and well-being outcomes. We aim to explore these challenges from the perspective of NEET youngsters and link workers who support them in co-creating individual action plans for their social and economic integration. Methods Data were collected through in-depth and semi-structured interviews with 15 young people in NEET situations, aged 18-34, and 7 link workers from several social organisations in rural and urban areas in Portugal. The young participants were selected based on their age, educational level, mental health needs, and urban/rural regions. Thematic analysis was used to analyse the interviews. Results The results showed a convergence between the perspectives of young people and link workers on the factors that influence social isolation and loneliness, including social and emotional skill deficits, excessive use of social media platforms, and mental health problems. Young people reported to being heavily involved in social media platforms, which often did not translate into meaningful relationships in real life. Youngsters also highlighted the lack of local support and community resources in rural areas. Link workers identified specific challenges young migrants face in community integration, including language barriers (for the youth with a migrant origin) and lack of familiarity with local resources. Conclusions This study highlights social media platforms, mental health and social and emotional skill deficits as main factors contributing to the social isolation and loneliness of young people in a NEET situation. This emphasizes the importance of integrated proximity-based interventions that incorporate community resource mapping and local support networks, including mental health support services, and promote multi-sector collaboration. Key messages • Social media platforms, mental health needs, and social and emotional skill deficits drive social isolation in NEET youth, stressing integrated, community-focused interventions. • Proximity-based strategies with community mapping and mental health support enhance NEET youths’ social integration, urging cross-sector collaboration.

High-resolution population mapping based on SDGSAT-1 glimmer imagery and deep learning: a case study of the Guangdong-Hong Kong-Marco Greater Bay Area

ABSTRACT Accurate population mapping is crucial for disaster management, urban planning, etc. However, current methods using nighttime light (NTL) and gridded population datasets are limited by low spatial resolution and insufficient training data for complex models such as deep learning. These models do not adequately utilize spatial information in population mapping. To address these limitations, this study proposes a high-resolution population mapping method using the Sustainable Development Goals Science Satellite 1 (SDGSAT-1) glimmer imager data and deep learning. The method includes a sample generation strategy with multiple regression and multilevel screening to provide sufficient, high-quality samples for deep learning. A Fine Population mapping network (FinePop-net) is also developed to train regression models using image samples, capturing multi-scale features for model training. When applied to the Guangdong-Hong Kong-Macao Greater Bay Area with 40-meter resolution SDGSAT 1 glimmer imagery, the method significantly reduced the average absolute error and root-mean-square error by 9.35% and 11.44%, respectively, compared with those of the pixel-level learning methods. It also outperformed other population spatialization datasets and NTL data by over 30% and 10%, respectively, in terms of error reduction. The results highlight the method’s effectiveness and the value of SDGSAT-1 glimmer imagery for fine population spatialization.

Transcriptome analyses of shell color and egg production traits between the uteruses of blue-green eggshell chickens and Hy-Line brown layers

Blue-green eggs exhibit unique shell color; however, compared to commercial layers, blue-green eggshell chickens have lower egg production and lack uniform shell colors. Aiming to confirm the molecular mechanisms that affect shell color and egg production, this study collected the uteruses of 12 blue-green eggshell chickens (BG group) and six Hy-Line layers (Brown group), which had significantly different shell color indexes (SCI) and egg numbers at 300 days of age (EN300). Transcriptome sequencing and comparative analyses were subsequently performed. BG hens were divided into two groups for comparative analysis (BGblue vs. BGgreen and BGlow vs. BGhigh, respectively), based on the differences in SCI and EN300, respectively. The result of weighted gene co-expression network (WGCNA) analysis showed that the sequenced and mapped 16,785 genes were clustered into 18 modules, among which six modules with a total of 4270 genes were highly correlated with SCI and EN300 traits. Five hundred and eleven differentially expressed genes (DEGs) belonged to the six key modules. Through KEGG mapping, GO enrichment, and Cytoscape network analysis, nine Hub genes were tightly associated with SCI and EN300. The up-regulated genes were CCR2, CCR8, CD40LG, IL18RAP, INHBA, and P2RY13, while the down-regulated genes were ABCA13, ADCY2, and GRM1. Co-analyses with the results of comparisons between the BG subgroups revealed that the expression of solute carrier (SLC) proteins and ABC transporters were highly related to eggshell color, while cytokine-cytokine receptor interactions and neuroactive ligand-receptor interactions were key pathways affecting egg production. The expression of extracellular cytokines and membrane receptors were significantly up-regulated in low-yield chickens. The candidate DEGs and pathways found in the study will assist in clarifying the molecular mechanisms affecting shell color and egg production, and improve the breeding of blue-green eggshell chickens.

Leveraging Digital Twins and Intrusion Detection Systems for Enhanced Security in IoT-Based Smart City Infrastructures

In this research, we investigate the integration of an Intrusion Detection System (IDS) with a Digital Twin (DT) to enhance the cybersecurity of physical devices in cyber–physical systems. Using Eclipse Ditto as the DT platform and Snort as the IDS, we developed a near-realistic test environment that included a Raspberry Pi as the physical device and a Kali Linux virtual machine to perform common cyberattacks such as Hping3 flood attacks and NMAP reconnaissance scans. The results demonstrated that the IDS effectively detected Hping3-based flood attacks but showed limitations in identifying NMAP scans, suggesting areas for IDS configuration improvements. Furthermore, the study uncovered significant system resource impacts, including high Central Processing Unit (CPU) usage during SYN and ACK flood attacks and persistent memory usage after Network Mapper (NMAP) scans, highlighting the need for enhanced recovery mechanisms. This research presents a novel approach by coupling a Digital Twin with an IDS, enabling real-time monitoring and providing a dual perspective on both system performance and security. The integration offers a holistic method for identifying vulnerabilities and understanding resource impacts during cyberattacks. The work contributes new insights into the use of Digital Twins for cybersecurity and paves the way for further research into automated defense mechanisms, real-world validation of the proposed model, and the incorporation of additional attack scenarios. The results suggest that this combined approach holds significant promise for enhancing the security and resilience of IoT devices and other cyber–physical systems.

Impact of multiplexing noise on multilayer networks of bistable maps

We explore numerically the complex dynamics of multilayer networks (consisting of three and one hundred layers) of cubic maps in the presence of noise-modulated interlayer coupling (multiplexing noise). The coupling strength is defined by independent discrete-time sources of color Gaussian noise. Uncoupled layers can demonstrate different complex structures, such as double-well chimeras, coherent and spatially incoherent regimes. Regions of partial synchronization of these structures are identified in the presence of multiplexing noise. We elucidate how synchronization of a three-layer network depends on the initially observed structures in the layers and construct synchronization regions in the plane of multiplexing noise parameters “noise spectrum width – noise intensity”. It is shown that in a hundred-layer network, clusters of synchronized layers can be formed at certain optimal values of multiplexing noise parameters. The performed numerical studies confidently indicate that the spatial dynamics of multilayer networks can be controlled by varying multiplexing noise parameters.

Localization and Network Connectivity of Lesions Causing Limb Ataxia in Patients With Stroke.

Ataxia is primarily considered to originate from the cerebellum. However, it can manifest without obvious cerebellar damage, such as in anterior circulation stroke, leaving the mechanisms of ataxia unclear. The aim of this study was to investigate whether stroke lesions causing limb ataxia localize to a common brain network. In this prospective cohort study, adult patients with new-onset stroke with visible lesions on CT or MRI from Turku University Hospital, Finland, were clinically examined (1) after their stroke while still admitted to the hospital (baseline) and (2) 4 months later (follow-up) to assess limb ataxia. Lesion locations and their functional connectivity, computed using openly available data from 1,000 healthy volunteers from the Brain Genome Superstruct Project, were compared voxel-by-voxel across the whole brain between patients with and without ataxia, using voxel-based lesion-symptom mapping and lesion network mapping. The findings were confirmed in an independent stroke patient cohort with identical clinical assessments. One hundred ninety-seven patients (mean age 67.2 years, 39%female) were included in this study. At baseline, 35 patients (68.3 years, 34%female) had and 162 (67.0 years, 40%female) did not have new-onset acute limb ataxia. At follow-up, additional 4 patients had developed late-onset limb ataxia, totalling to 39 patients (68.6 years, 36%female) with limb ataxia at any point. One hundred eighteen patients (66.2 years, 40%female) did not have ataxia at any point (n = 40 with missing follow-up data). Lesions in 54% of the patients with acute limb ataxia were located outside the cerebellum and cerebellar peduncles, and we did not find an association between specific lesion locations and ataxia. Lesions causing acute limb ataxia, however, were connected to a common network centered on the intermediate zone cerebellum and cerebellar peduncles (lesion connectivity in patients with vs without acute limb ataxia, pFWE < 0.05). The results were similar when comparing patients with and without ataxia at any point, and when excluding lesions in the cerebellum and cerebellar peduncles (pFWE < 0.05). The findings were confirmed in the independent stroke dataset (n = 96), demonstrating an OR of 2.27 (95% CI 1.32-3.91) for limb ataxia per standard deviation increase in limb ataxia network damage score. Lesions causing limb ataxia occur in heterogeneous locations but localize to a common brain network.

Mapping connectivity and network alterations in youth depression by functional brain imaging

Mapping connectivity and network alterations in youth depression by functional brain imaging

Super-resolution water body mapping with a feature collaborative CNN model by fusing Sentinel-1 and Sentinel-2 images

Mapping water bodies from remotely sensed imagery is crucial for understanding hydrological and biogeochemical processes. The identification of water extent is mainly dependent on optical and synthetic aperture radar (SAR) images. However, the use of remote sensing for water body mapping is often undermined by the mixed pixel dilemma inherent to traditional hard classification approaches. At the same time, the presence of clouds in optical imagery and speckle noise in SAR imagery, coupled with the difficulty in differentiating between water-like surfaces and actual water bodies, significantly compromise the accuracy of water body identification. This paper proposes a DEEP feature collaborative convolutional neural network (CNN) for Water Super-Resolution Mapping based on Optical and SAR images (DeepOSWSRM), which collaboratively leverages Sentinel-1 and Sentinel-2 imagery to address the challenges of missing data and mixed pixels. The Sentinel-1 image provides complementary water distribution information for the cloudy areas of the Sentinel-2 image, while the Sentinel-2 image enhances the perception capabilities for small water bodies in the Sentinel-1 image. Using PlanetScope imagery as the true reference data, the effectiveness of the proposed method was assessed through two experimental scenarios: one utilizing synthetic coarse-resolution imagery degraded from Sentinel-1 and Sentinel-2 data and another using actual Sentinel-1 and Sentinel-2 data, encompassing both simulated and real cloud conditions. A comparative analysis was conducted against three state-of-the-art CNN-based water mapping methods and two CNN SRM methods. The findings demonstrate that the proposed DeepOSWSRM method successfully produces accurate, fine-resolution water body maps, with its performance mainly benefiting from the fusion of SAR and optical images.

DMMGNet: A discrimination mapping and memory bank mean guidance-based network for high-performance few-shot industrial anomaly detection

For deep learning-based industrial anomaly detection, it is still challenging to get adequate images for model training and achieve cold start for cross-product migration, restricting their practical application in real industrial production. Herein, an innovative few-shot anomaly detection network DMMGNet based on discrimination mapping and memory bank mean guidance strategies are demonstrated, which is trained by a new two-branch data augmentation technique. By separating the features stored in memory bank from the features used for training, the two-branch data augmentation method can significantly improve the robustness of few-shot model training and reduce the redundance of memory bank. In the elaborately designed discrimination mapping module, new negative samples are generated by adding dynamic Gaussian noise to normal samples along the channel dimension in feature space to solve the problem of sample imbalance. Meanwhile, the discrimination mapping module also helps to map the feature distribution of positive samples to the target domain more efficiently and reduce the deviation of feature domain, conducive to a more precise separation of positive and negative samples. In addition, a novel mean guidance approach with an optimized loss function is developed to guide the positive sample feature mapping by specifying the local feature space center to form a clear feature domain contour and enhance the detection accuracy. The multiple experimental results validate that our DMMGNet outperforms the most advanced anomaly detection counterparts on image-level AUROC, showing an increase by 0.3–3 % on both MVTec AD and MPDD benchmarks under several few-shot scenarios.

Using a Dual-Disease Target Mapping Network Pharmacology Approach, Verbascoside Ameliorates Osteoporosis by Activating Estrogen Signaling to Alleviate Oxidative Stress.

Verbascoside, a compound classified as a phenylethanol glycoside in Dihuang, has been the subject of modern pharmacological investigations. These studies have revealed its noteworthy antioxidant, anti-inflammatory, memory-enhancing, neuroprotective, antitumor, and various other pharmacological properties. While verbascoside exhibits favorable antioxidant effects, its precise mechanism of action in ameliorating osteoporosis through the treatment of oxidative stress remains unclear. This study employed CCK8, ALP, ELISA, and ROS staining techniques to examine the osteoporotic effects of verbascoside on zebrafish and MC3T3-E1 cells. Additionally, this study aimed to investigate the molecular mechanism by which verbascoside improves osteoporosis by mitigating oxidative stress. To identify the common targets of verbascoside in relation to oxidative stress and osteoporosis, network pharmacology and molecular dynamics simulation were employed. The construction of the verbascoside - oxidative stress - osteoporosis - potential target gene network aimed to identify the core targets, while the mechanism of action was elucidated through KEGG analysis, and the accuracy was confirmed by assessing the mRNA expression of the targets. In vivo experiments demonstrated that verbascoside exhibited therapeutic effects on osteoporosis and reduced ROS production in zebrafish. In vitro experiments further revealed that verbascoside enhanced the proliferation and differentiation of MC3T3-E1 cells, thereby improving the oxidative stress status of osteoblasts. Thirteen core targets and estrogen signaling pathways were identified through the application of network pharmacology. The pivotal role of the estrogen signaling pathway in facilitating the ability of verbascoside to mitigate oxidative stressinduced osteoporosis was substantiated by the modulation of target protein mRNA expression. The findings underscore the considerable therapeutic potential of verbascoside in ameliorating osteoporosis through the alleviation of oxidative stress, thus establishing it as a promising compound for the treatment of this condition.

Privacy-SF: An encoding-based privacy-preserving segmentation framework for medical images

Deep learning is becoming increasingly popular and is being extensively used in the field of medical image analysis. However, the privacy sensitivity of medical data limits the availability of data, which constrains the advancement of medical image analysis and impedes collaboration across multiple centers. To address this problem, we propose a novel encoding-based framework, named Privacy-SF, aimed at implementing privacy-preserving segmentation for medical images. Our proposed segmentation framework consists of three CNN networks: 1) two encoding networks on the client side that encode medical images and their corresponding segmentation masks individually to remove the privacy features, 2) a unique mapping network that analyzes the content of encoded data and learns the mapping from the encoded image to the encoded mask. By sequentially encoding data and optimizing the mapping network, our approach ensures privacy protection for images and masks during both the training and inference phases of medical image analysis. Additionally, to further improve the segmentation performance, we carefully design augmentation strategies specifically for encoded data based on its sequence nature. Extensive experiments conducted on five datasets with different modalities demonstrate excellent performance in privacy-preserving segmentation and multi-center collaboration. Furthermore, the analysis of encoded data and the experiment of model inversion attacks validate the privacy-preserving capability of our approach.

A lightweight CNN-Transformer network for pixel-based crop mapping using time-series Sentinel-2 imagery

Deep learning approaches have provided state-of-the-art performance in crop mapping. Recently, several studies have combined the strengths of two dominant deep learning architectures, Convolutional Neural Networks (CNNs) and Transformers, to classify crops using remote sensing images. Despite their success, many of these models utilize patch-based methods that require extensive data labeling, as each sample contains multiple pixels with corresponding labels. This leads to higher costs in data preparation and processing. Moreover, previous methods rarely considered the impact of missing values caused by clouds and no-observations in remote sensing data. Therefore, this study proposes a lightweight multi-stage CNN-Transformer network (MCTNet) for pixel-based crop mapping using time-series Sentinel-2 imagery. MCTNet consists of several successive modules, each containing a CNN sub-module and a Transformer sub-module to extract important features from the images, respectively. An attention-based learnable positional encoding (ALPE) module is designed in the Transformer sub-module to capture the complex temporal relations in the time-series data with different missing rates. Arkansas and California in the U.S. are selected to evaluate the model. Experimental results show that the MCTNet has a lightweight advantage with the fewest parameters and memory usage while achieving the superior performance compared to eight advanced models. Specifically, MCTNet obtained an overall accuracy (OA) of 0.968, a kappa coefficient (Kappa) of 0.951, and a macro-averaged F1 score (F1) of 0.933 in Arkansas, and an OA of 0.852, a Kappa of 0.806, and an F1 score of 0.829 in California. The results highlight the importance of each component of the model, particularly the ALPE module, which enhanced the Kappa of MCTNet by 4.2% in Arkansas and improved the model’s robustness to missing values in remote sensing data. Additionally, visualization results demonstrated that the features extracted from CNN and Transformer sub-modules are complementary, explaining the effectiveness of the MCTNet.

Research on low-power driving fatigue monitoring method based on spiking neural network.

Fatigue driving is one of the leading causes of traffic accidents, and the rapid and accurate detection of driver fatigue is of paramount importance for enhancing road safety. However, the application of deep learning models in fatigue driving detection has long been constrained by high computational costs and power consumption. To address this issue, this study proposes an approach that combines Self-Organizing Map (SOM) and Spiking Neural Networks (SNN) to develop a low-power model capable of accurately recognizing the driver's mental state. Initially, spatial features are extracted from electroencephalogram (EEG) signals using the SOM network. Subsequently, the extracted weight vectors are encoded and fed into the SNN for fatigue driving classification. The research results demonstrate that the proposed method effectively considers the spatiotemporal characteristics of EEG signals, achieving efficient fatigue detection. Simultaneously, this approach successfully reduces the model's power consumption. When compared to traditional artificial neural networks, our method reduces energy consumption by approximately 12.21-42.59%.

Hierarchical multi-level dynamic hyperparameter deformable image registration with convolutional neural network

Objective. To enable the registration network to be trained only once, achieving fast regularization hyperparameter selection during the inference phase, and to improve registration accuracy and deformation field regularity. Approach. Hyperparameter tuning is an essential process for deep learning deformable image registration (DLDIR). Most DLDIR methods usually perform a large number of independent experiments to select the appropriate regularization hyperparameters, which are time-consuming and resource-consuming. To address this issue, we propose a novel dynamic hyperparameter block, which comprises a distributed mapping network, dynamic convolution, attention feature extraction layer, and instance normalization layer. The dynamic hyperparameter block encodes the input feature vectors and regularization hyperparameters into learnable feature variables and dynamic convolution parameters which changes the feature statistics of the high-dimensional features layer feature variables, respectively. In addition, the proposed method replaced the single-level structure residual blocks in LapIRN with a hierarchical multi-level architecture for the dynamic hyperparameter block in order to improve registration performance. Main results. On the OASIS dataset, the proposed method reduced the percentage of |Jϕ|⩽0 by 28.01 % , 9.78 % and improved Dice similarity coefficient by 1.17 % , 1.17 % , compared with LapIRN and CIR, respectively. On the DIR-Lab dataset, the proposed method reduced the percentage of |Jϕ|⩽0 by 10.00 % , 5.70 % and reduced target registration error by 10.84 % , 10.05 % , compared with LapIRN and CIR, respectively. Significance. The proposed method can fast achieve the corresponding registration deformation field for arbitrary hyperparameter value during the inference phase. Extensive experiments demonstrate that the proposed method reduces training time compared to DLDIR with fixed regularization hyperparameters while outperforming the state-of-the-art registration methods concerning registration accuracy and deformation smoothness on brain dataset OASIS and lung dataset DIR-Lab.

Time-series China urban land use mapping (2016–2022): An approach for achieving spatial-consistency and semantic-transition rationality in temporal domain

The global urbanization trend is geographically manifested through city expansion and the renewal of internal urban structures and functions. Time-series urban land use (ULU) maps are vital for capturing dynamic land changes in the urbanization process, giving valuable insights into urban development and its environmental consequences. Recent studies have mapped ULU in some cities with a unified model, but ignored the regional differences among cities; and they generated ULU maps year by year, but ignored temporal correlations between years; thus, they could be weak in large-scale and long time-series ULU monitoring. Accordingly, we introduce an temporal-spatial-semantic collaborative (TSS) mapping framework to generating accurate ULU maps with considering regional differences and temporal correlations. Firstly, to support model training, a large-scale ULU sample dataset based on OpenStreetMap (OSM) and Sentinel-2 imagery is automatically constructed, providing a total number of 56,412 samples with a size of 512 × 512 which are divided into six sub-regions in China and used for training different classification models. Then, an urban land use mapping network (ULUNet) is proposed to recognize ULU. This model utilizes a primary and an auxiliary encoder to process noisy OSM samples and can enhance the model's robustness under noisy labels. Finally, taking the temporal correlations of ULU into consideration, the recognized ULU are optimized, whose boundaries are unified by a time-series co-segmentation, and whose categories are modified by a knowledge-data driven method. To verify the effectiveness of the proposed method, we consider all urban areas in China (254,566 km2), and produce a time-series China urban land use dataset (CULU) at a 10-m resolution, spanning from 2016 to 2022, with an overall accuracy of CULU is 82.42%. Through comparison, it can be found that CULU outperforms existing datasets such as EULUC-China and UFZ-31cities in data accuracies, spatial boundaries consistencies and land use transitions logicality. The results indicate that the proposed method and generated dataset can play important roles in land use change monitoring, ecological-environmental evolution analysis, and also sustainable city development.

A Novel Sample Generation Method for Deep Learning Lithological Mapping with Airborne TASI Hyperspectral Data in Northern Liuyuan, Gansu, China

High-resolution and thermal infrared hyperspectral data acquired from the Thermal Infrared Airborne Spectrographic Imager (TASI) have been recognized as efficient tools in geology, demonstrating significant potential for rock discernment. Deep learning (DL), as an advanced technology, has driven substantial advancements in lithological mapping by automatically extracting high-level semantic features from images to enhance recognition accuracy. However, gathering sufficient high-quality lithological samples for model training is challenging in many scenarios, posing limitations for data-driven DL approaches. Moreover, existing sample collection approaches are plagued by limited verifiability, subjective bias, and variation in the spectra of the same class at different locations. To tackle these challenges, a novel sample generation method called multi-lithology spectra sample selection (MLS3) is first employed. This method involves multiple steps: multiple spectra extraction, spectra combination and optimization, lithological type identification, and sample selection. In this study, the TASI hyperspectral data collected from the Liuyuan area in Gansu Province, China, were used as experimental data. Samples generated based on MLS3 were fed into five typical DL models, including two-dimensional convolutional neural network (2D-CNN), hybrid spectral CNN (HybridSN), multiscale residual network (MSRN), spectral-spatial residual network (SSRN), and spectral partitioning residual network (SPRN) for lithological mapping. Among these models, the accuracy of the SPRN reaches 84.03%, outperforming the other algorithms. Furthermore, MLS3 demonstrates superior performance, achieving an overall accuracy of 2.25–6.96% higher than other sample collection methods when SPRN is used as the DL framework. In general, MLS3 enables both the quantity and quality of samples, providing inspiration for the application of DL to hyperspectral lithological mapping.

Training and Validating a Neural Network for Myocardial Blood Flow Mapping in 82Rb PET: A Multicenter Study

Training and Validating a Neural Network for Myocardial Blood Flow Mapping in 82Rb PET: A Multicenter Study

An Embedded Neural Network Approach for Reinforcing Deep Learning: Advancing Hand Gesture Recognition

Deep neural networks (DNNs) can face limitations during training for recognition, motivating this study to improve recognition capabilities by optimizing deep learning features for hand gesture image recognition. We propose a novel approach that enhances features from well-trained DNNs using an improved radial basis function (RBF) neural network, targeting recognition within individual gesture categories. We achieve this by clustering images with a self-organizing map (SOM) network to identify optimal centers for RBF training. Our enhanced SOM, employing the Hassanat distance metric, outperforms the traditional K-Means method across a comparative analysis of various distance functions and the expanded number of cluster centers, accurately identifying hand gestures in images. Our training pipeline learns from hand gesture videos and static images, addressing the growing need for machines to interact with gestures. Despite challenges posed by gesture videos, such as sensitivity to hand pose sequences within a single gesture category and overlapping hand poses due to the high similarities and repetitions, our pipeline achieved significant enhancement without requiring time-related training data. We also improve the recognition of static hand pose images within the same category. Our work advances DNNs by integrating deep learning features and incorporating SOM for RBF training.

Deep multi-sphere support vector data description based on disentangled representation learning

Deep support vector data description (Deep SVDD) combines deep mapping network and support vector data description (SVDD) to jointly optimize network connection weights and hypersphere volume. However, when the parameters of deep mapping network are set improperly, Deep SVDD may face the problem of hypersphere collapse, where all input data are mapped as the hypersphere center. To overcome the hypersphere collapse problem of Deep SVDD and improve the feature learning ability of deep mapping network, deep multi-sphere SVDD based on disentangled representation learning (DMSVDD-DRL) is proposed. DMSVDD-DRL consists of a variational autoencoder (VAE) and multiple hyperspheres. The feature representations obtained by VAE are disentangled into discriminative representations and generative representations that obey mixture t-distribution and Gaussian distribution, respectively. In the pre-training phase of DMSVDD-DRL, the network parameters and the hypersphere centers are initialized. In the training phase, the augmented data are added into the training set. The discriminative representations of both the input and augmented data are generated through the mapping network. Furthermore, multiple hyperspheres are constructed by the obtained discriminative representations in the feature space. Finally, the VAE loss of the input data, the reconstruction error of the augmented data, the augmentation loss between the input and augmented data, the average radius of the multiple hyperspheres, and the average distance from discriminative representations to their corresponding hypersphere centers are jointly minimized to obtain the optimal network connection weights and the multiple minimum volume hyperspheres. The effectiveness of the proposed DMSVDD-DRL is validated through the comparative and ablation experiments on the benchmark data sets. In addition, it is verified that DMSVDD-DRL is more robust against outliers in comparison with its related methods.