Transfer Network Research Articles

Privacy-Preserving Cross-Area Traffic Forecasting in ITS: A Transferable Spatial-Temporal Graph Neural Network Approach

Traffic forecasting is essential in improving and maintaining safety and orderliness in intelligent transportation systems (ITS). As a deep learning approach, graph neural networks (GNN) based spatial-temporal association mining methods are promising in traffic forecasting. However, current GNN-based methods usually require a high number of training data, and when the sample volume is small, the performance of the model drops dramatically. The existing transfer methods can solve this problem by leveraging knowledge from other data-rich areas, but the domain adaption method with access to source data still faces the non-neglectable problem of private information leakage in the source area. A solution that can solve cross-area transfer without access to source data is still missing. In this paper, to fill the gap, we propose a Transferable Federated Inductive Spatial-Temporal Graph Neural Network (T-ISTGNN) framework to transfer spatial-temporal dependency information in cross-area data to accomplish traffic state forecasting. First, we introduce a multi-source model aggregation scheme based on federated learning to retain the traffic information of the source areas. Second, we propose a transfer method between source and target areas based on hypothesis transfer learning to achieve domain adaption under source domain data protection. Third, we propose a GNN-based method called Inductive Spatial-Temporal Graph Neural Network (ISTGNN) for traffic forecasting. Experiments on real-world datasets demonstrate that T-ISTGNN is capable of cross-area traffic state forecasting under the restriction of preserving the privacy of source areas.

Transferable 3C-SiC nanowire network film for high performance flexible UV photodetectors with a simple, low cost and large scale production

Transferable 3C-SiC nanowire network film for high performance flexible UV photodetectors with a simple, low cost and large scale production

Multi-representation transferable attention network for remaining useful life prediction of rolling bearings under multiple working conditions

Remaining useful life (RUL) prediction improves equipment reliability by avoiding unexpected breakdowns. However, different operating conditions may cause the performance degradation of the prediction model due to the domain-shift problem. Therefore, a novel method based on the multiple representation transferable attention network (MRTAN) is proposed in this paper for RUL prediction under multiple working conditions. In the MRTAN-based method, multiple convolutional modules are first used to learn the degradation features. Next, the multi-representation adaptation module is utilized to further mine the multiple domain-invariant representations. Then, during the training process, the transferable attention module can activate the representation with high transferability dynamically. Finally, the prognostic model is optimized by multiple optimization objectives and the back-propagation algorithm. Besides, multiple cross-domain RUL prediction tasks are employed to validate the effectiveness of the MRTAN-based model. Experiments demonstrate that the proposed method can provide better prognostic performance and avoid the negative transfer problem.

Implementation of Cellular-Based Drone Module using Cloud Services

<div class="page" title="Page 1"><div class="layoutArea"><div class="column"><p>The notion of a smart city incorporates the integration of infrastructure, services, and the community and encompasses the deployment of unmanned aerial vehicles (UAVs) for monitoring crop fields, facilitating logistics delivery, and performing high-altitude cleaning tasks. In a smart city, the interconnectedness of devices is realized through the medium of the Internet of Things (IoT). This research endeavors to explore the usage of Beyond Visual Line of Sight (BVLOS) for enabling remote command and control of UAV/UGV modes, leveraging 4G/LTE connectivity as an enabler. 4G/LTE connectivity is known for its improvement in data transfer speed and network capacity, which potentially enables the connection of more devices, including drones. The high availability and scalability of cloud services have become crucial factors in utilizing cloud services as the most cost-effective and expedient relay for connecting two nodes over the internet globally for now. The proposed methodology would be integrated into a smart drone module, which would be deployed at a small scale as a component of the Intelligent Transportation System (ITS).<br /><br /></p></div></div></div>

Designing continuous proton-conductive channels for direct methanol fuel cell through the sulfonated poly(ether ether ketone)/carbon quantum dot/graphitic carbon nitride nanosheet

In this research, new nanocomposite membranes were fabricated based on sulfonated poly(ether ether ketone)/carbon quantum dot/graphitic carbon nitride nanosheet (SPEEK/CQD/g-C3N4). It was found that anchoring CQD onto g-C3N4 provides a new proton transport pathway in the design of proton exchange membranes for high-performance fuel cell applications. The uniform distribution of CQDs on the nanosheets and the interaction of hydrophilic groups such as the –NH2 and CO with the -SO3H of the polymer matrix creates a hydrogen bond transfer network and facilitates the protons transfer based on the Grotthuss mechanism. The incorporation of CQD/g-C3N4 in the polymer matrix has increased the proton conduction from 0.023 S/cm for SPEEK at 20 °C and 100 % R.H. to 0.052 S/cm for the SPEEK/CQD/g-C3N4-3 %wt. The proton conduction for the nanocomposite membrane containing 3 wt% CQD/g-C3N4 improved at higher temperatures (0.128 S/cm at 80 °C) and also, this membrane showed better thermal properties (maximum Tg at 176 °C). Compared to the SPEEK membrane, the nanocomposite membrane with 3 wt% CQD/g-C3N4 had a better single-cell direct methanol fuel cell (DMFC) test performance.

Fusing joint distribution and adversarial networks: A new transfer learning method for intelligent fault diagnosis

As integral components within rotating machinery, bearings and gears pose a critical challenge in fault diagnosis. Presently, data-driven fault diagnosis stands out as a viable approach. However, real-world operational variations readily induce domain shifts, complicating transfer learning and diminishing the diagnostic efficacy of models. The process of re-labeling the fault categories of model demands substantial time and financial resources. Consequently, to surmount these challenges, this study introduces a novel unsupervised transfer learning framework that leverages the amalgamation of joint distribution and adversarial networks for diagnosing faults in bearings and gears within rotating machinery.The joint adaptation network facilitates the learning of the transfer network by aligning the joint distribution across multiple specific domain layers. This alignment is achieved through the application of joint maximum mean discrepancy (JMMD) within the joint network. Simultaneously, the adversarial network employs a domain classifier to minimize the domain classification loss, treating it as the difference in domain distribution to mitigate domain shift effectively. The integration of these two methodologies accomplishes domain alignment, reduces model training time, and enhances the accuracy and stability of the diagnostic model.Validation of the proposed model framework is conducted using four sets of bearing faults and six sets of gear faults. The results confirm the superior accuracy and stability of the new model framework in addressing bearing and gear faults within the realm of rotating machinery.

Synergistic behavior of organic and inorganic constituents in improving the charge transfer ability of Polyoxometalate-Based ionic liquids

The electron transfer kinetics and electronic properties of (3-sulfonic acid) propyl bipyridine-12tungstophosphoric acid (BPyPS-PW) as a novel organic–inorganic ionic liquid have been investigated. Cyclic voltammetry (CV) revealed that BPyPS enhanced the oxidation peak currents of tungstophosphoric acid (PW) by 1.4-fold. According to linear sweep voltammetry (LSV), BPyPS-PW exhibited significantly lower tafel slope (96 mV) than BPyPS and PW (144 and 119 mV, respectively). This indicates that the current density increases more rapidly with overpotential, pointing the superior electron transfer ability of BPyPS-PW. Further, it showed about 1.5- and 1.2-fold higher charge transfer coefficient than BPyPS and PW. Differential pulse voltammetry (DPV) and square wave voltammetry (SWV) showed increased peak current intensities for the hybrid compared to both constituents. Electrochemical impedance spectroscopy (EIS) indicated that BPyPS-PW has the lowest charge transfer resistance (RCT, 685 Ω), Warburg impedance (ZW, 1.87 Ω), electrolyte solution resistance (RS, 86 Ω), and double-layer capacitance resistance (ZC, 38.73 Ω) among the studied compounds, indicative of its reduced electron transfer resistance at the electrode/electrolyte interface. According to chronoamperometry (CA), BPyPS-PW exhibited higher electrostability, with only 21.48 % current decay after 120 s, while BPyPS and PW experienced a stronger decay (50.09 % and 24.57 %, respectively), attributed to a synergistic behavior of the constituents that results in a larger charge transfer network and more electron active centers. These findings are crucial for the design of new ionic liquid electrolytes to be used in electrochemical applications.

Dynamic model-assisted transferable network for liquid rocket engine fault diagnosis using limited fault samples

The accurate detection and diagnosis of faults in Liquid Rocket Engines (LREs) are critical for ensuring space mission safety. However, the limited availability of actual fault samples and the diversification of potential faults present significant challenges in achieving precise diagnosis. To overcome these obstacles, we propose a dynamic model-assisted transfer learning approach. In this study, we first modularize the LRE and establish a dynamic model based on the mathematical principles of each module. Subsequently, we artificially established a fault module model, injected faults into the normal model, and simulated various fault modes to expand the fault sample library. Leveraging this augmented dataset in combination with a limited number of actual fault samples, we employ transfer learning to fine-tune a Convolutional Neural Network (CNN). Compared with other classic methods, the migrated CNN effectively adapts to the distribution of real data and significantly improves the accuracy of LRE fault diagnosis.

Remote Sensing Image Super-Resolution via Multi-Scale Texture Transfer Network

As the degradation factors of remote sensing images become increasingly complex, it becomes challenging to infer the high-frequency details of remote sensing images compared to ordinary digital photographs. For super-resolution (SR) tasks, existing deep learning-based single remote sensing image SR methods tend to rely on texture information, leading to various limitations. To fill this gap, we propose a remote sensing image SR algorithm based on a multi-scale texture transfer network (MTTN). The proposed MTTN enhances the texture feature information of reconstructed images by adaptively transferring texture information according to the texture similarity of the reference image. The proposed method adopts a multi-scale texture-matching strategy, which promotes the transmission of multi-scale texture information of remote sensing images and obtains finer-texture information from more relevant semantic modules. Experimental results show that the proposed method outperforms state-of-the-art SR techniques on the Kaggle open-source remote sensing dataset from both quantitative and qualitative perspectives.

SSAT ++ : A Semantic-Aware and Versatile Makeup Transfer Network With Local Color Consistency Constraint.

The purpose of makeup transfer (MT) is to transfer makeup from a reference image to a target face while preserving the target's content. Existing methods have made remarkable progress in generating realistic results but do not perform well in terms of semantic correspondence and color fidelity. In addition, the straightforward extension of processing videos frame by frame tends to produce flickering results in most methods. These limitations restrict the applicability of previous methods in real-world scenarios. To address these issues, we propose a symmetric semantic-aware transfer network (SSAT ++ ) to improve makeup similarity and video temporal consistency. For MT, the feature fusion (FF) module first integrates the content and semantic features of the input images, producing multiscale fusion features. Then, the semantic correspondence from the reference to the target is obtained by measuring the correlation of fusion features at each position. According to semantic correspondence, the symmetric mask semantic transfer (SMST) module aligns the reference makeup features with the target content features to generate MT results. Meanwhile, the semantic correspondence from the target to the reference is obtained by transposing the correlation matrix and applied to the makeup removal task. To enhance color fidelity, we propose a novel local color loss that forces the transferred results to have the same color histogram distribution as the reference. Furthermore, a morphing simulation is designed to ensure temporal consistency for video MT without requiring additional video frame input and optical flow estimation. To evaluate the effectiveness of our SSAT ++ , extensive experiments have been conducted on the MT dataset which has a variety of makeup styles, and on the MT-Wild dataset which contains images with diverse poses and expressions. The experiments show that SSAT ++ outperforms existing MT methods through qualitative and quantitative evaluation and provides more flexible makeup control. Code and trained model will be available at https://gitee.com/sunzhaoyang0304/ssat-msp and https://github.com/Snowfallingplum/SSAT.

3D composite carbon Foam-Loaded MnO2 and FeOOH enabling the “oxidation-adsorption” effect toward efficient removal of arsenic from groundwater

The remediation of geogenic high arsenic groundwater is of great significance for environmental protection. In this study, a feasible strategy for in-situ growth of MnO2 and FeOOH on a three-dimensional (3D) porous carbon foam skeleton was developed by simple calcination carbonization and activation treatment of modified carbon foams. This work demonstrates that arsenic in aqueous solution can be efficiently removed by the dual function of “oxidation and adsorption” from the composite of MnO2-FeOOH co-loaded carbon foam (CF). The prepared 3D monolithic has the advantages of low cost, easy preparation, and independent integrity. It can be innovatively used for arsenic remediation of groundwater systems. The MnO2 serves as the oxidizer to convert highly toxic As(III) into less toxic As(V), the FeOOH acts as the adsorbent for the capture of As(V), and CF offers an interconnected porous network for efficient mass transfer as well as a robust skeleton for facile recycling. The maximum adsorption capacity of As(III) and As(V) for MnO2-FeOOH@CF can reach 63 and 78 mg g−1, respectively. After repeatedly utilizing the free-standing adsorbent three times, its adsorption capacity for As(III) and As(V) can still reach 84 % and 85 % of the initial adsorption capacity, respectively. The low cost, high efficiency of arsenic removal, and easy preparation and recycling endow the MnO2-FeOOH@CF to be a promising adsorbent in practical application.

Amphibious epidermal area networks for uninterrupted wireless data and power transfer

The human body exhibits complex, spatially distributed chemo-electro-mechanical processes that must be properly captured for emerging applications in virtual/augmented reality, precision health, activity monitoring, bionics, and more. A key factor in enabling such applications involves the seamless integration of multipurpose wearable sensors across the human body in different environments, spanning from indoor settings to outdoor landscapes. Here, we report a versatile epidermal body area network ecosystem that enables wireless power and data transmission to and from battery-free wearable sensors with continuous functionality from dry to underwater settings. This is achieved through an artificial near field propagation across the chain of biocompatible, magneto-inductive metamaterials in the form of stretchable waterborne skin patches—these are fully compatible with pre-existing consumer electronics. Our approach offers uninterrupted, self-powered communication for human status monitoring in harsh environments where traditional wireless solutions (such as Bluetooth, Wi-Fi or cellular) are unable to communicate reliably.

Novel adaptive loss weighted transfer network for partial domain fault diagnosis

Mechanical fault transfer diagnosis has been confirmed as a feasible approach for tackling intelligent diagnosis with incomplete fault information and scarce labeled data on the basis of big data through the transfer of diagnostic knowledge from one or more conditions to any other condition. However, existing research has developed a hypothesis, i.e., the target domain shares an identical label space with the source domain, making it unfeasible to address the practical issue that the target domain label space is a subset of the source domain label space, resulting in low transfer diagnosis accuracy. To address this issue, a novel unsupervised intelligent diagnosis approach named double classifiers-dependent transfer diagnosis network is developed. In this approach, the label distribution weights are generated through the probability output of the classifier of source domain label space to target domain samples, by which small weights are assigned to irrelevant source samples to avoid negative transfer in the global-local maximum mean discrepancies (GL-MMD). In addition, classifiers of the source domain label space and the shared label space are built separately to improve the reliability of label distribution weights and GL-MMD. By training the network in the shared label space, diagnostic knowledge in partial domain issues is effectively transferred. Two cases are implemented to verify the effectiveness of the developed approach. Compared with other transfer diagnosis approaches, the developed approach achieved better diagnostic performance.

Machinery Prognostics and High-Dimensional Data Feature Extraction Based on a Transformer Self-Attention Transfer Network.

Machinery degradation assessment can offer meaningful prognosis and health management information. Although numerous machine prediction models based on artificial intelligence have emerged in recent years, they still face a series of challenges: (1) Many models continue to rely on manual feature extraction. (2) Deep learning models still struggle with long sequence prediction tasks. (3) Health indicators are inefficient for remaining useful life (RUL) prediction with cross-operational environments when dealing with high-dimensional datasets as inputs. This research proposes a health indicator construction methodology based on a transformer self-attention transfer network (TSTN). This methodology can directly deal with the high-dimensional raw dataset and keep all the information without missing when the signals are taken as the input of the diagnosis and prognosis model. First, we design an encoder with a long-term and short-term self-attention mechanism to capture crucial time-varying information from a high-dimensional dataset. Second, we propose an estimator that can map the embedding from the encoder output to the estimated degradation trends. Then, we present a domain discriminator to extract invariant features from different machine operating conditions. Case studies were carried out using the FEMTO-ST bearing dataset, and the Monte Carlo method was employed for RUL prediction during the degradation process. When compared to other established techniques such as the RNN-based RUL prediction method, convolutional LSTM network, Bi-directional LSTM network with attention mechanism, and the traditional RUL prediction method based on vibration frequency anomaly detection and survival time ratio, our proposed TSTN method demonstrates superior RUL prediction accuracy with a notable SCORE of 0.4017. These results underscore the significant advantages and potential of the TSTN approach over other state-of-the-art techniques.

A cascaded framework with cross-modality transfer learning for whole heart segmentation

Automatic and accurate segmentation of the whole heart structure from 3D cardiac images plays an important role in helping physicians diagnose and treat cardiovascular disease. However, the time-consuming and laborious manual labeling of the heart images results in the inefficiency of utilizing the existing CT or MRI for training the deep learning network, which decrease the accuracy of whole heart segmentation. However, multi-modality data contains multi-level information of cardiac images due to different imaging mechanisms, which is beneficial to improve the segmentation accuracy. Therefore, this paper proposes a cascaded framework with cross-modality transfer learning for whole heart segmentation (CM-TranCaF), which consists of three key modules: modality transfer network (MTN), U-shaped multi-attention network (MAUNet) and spatial configuration network (SCN). In MTN, MRI images are transferred from MRI domain to CT domain, to increase the data volume by adopting the idea of adversarial training. The MAUNet is designed based on UNet, while the attention gates (AGs) are integrated into the skip connection to reduce the weight of background pixels. Moreover, to solve the problem of boundary blur, the position attention block (PAB) is also integrated into the bottom layer to aggregate similar features. Finally, the SCN is used to finetune the segmentation results by utilizing the anatomical information between different cardiac substructures. By evaluating the proposed method on the dataset of the MM-WHS challenge, CM-TranCaF achieves a Dice score of 91.1% on the testing dataset. The extensive experimental results prove the effectiveness of the proposed method compared to other state-of-the-art methods.

Complex network-based analysis of inter-basin water transfer networks

Driven by water resources scarcity, inter-basin water transfer (IBWT) projects have emerged and evolved, thereby shaping the current water supply framework. As IBWT projects increase, they gradually acquire network characteristics, giving rise to IBWT networks. However, the impact of the development of IBWT networks on water-supply patterns remains unclear? Analyzing the spatial and temporal growth patterns of IBWT networks through the lens of complex network theory can help answer this question. This study established a framework for analyzing IBWT networks using complex network analysis (CNA). Within this framework, we analyzed the spatial and temporal development characteristics of the IBWT network by quantifying its global properties across three dimensions: efficiency, elasticity, and coordination. Six centrality indicators were employed to measure the significance of individual nodes. Finally, the Infomap method was employed for network community detection. The results demonstrated that the CNA-based framework effectively captured the comprehensive development of the IBWT network, which underwent six stages from inception to high-speed, near-term, and long-term Stages. IBWT network efficiency, elasticity, and coordination all showed growth, which was most pronounced in 2013–2022, where the scaled global efficiency, network efficiency, average node connectivity, and average betweenness centrality metrics all grew by a factor of greater than 4. Importantly, the total percentage of centrality greater than 0.2 in the Huang-Huai-Hai-Chang Basin was 63.0%. The 125 initial IBWT projects in the IBWT network contained 58 communities, while the subsequent construction of 57 projects added only 12 communities. Future new IBWT network projects will tend to joined existing communities.

Make a song curative: A spatio-temporal therapeutic music transfer model for anxiety reduction

Anxiety is a prevalent mental health issue that affects both individuals and society. Music therapy has been found effective in reducing anxiety in clinical trials. However, the available music pieces for therapy are limited and may not meet the requirements of individuals from diverse backgrounds. To overcome this limitation, we propose a novel model called the spatio-temporal therapeutic music transfer model (StTMTM) for anxiety reduction. The StTMTM can transfer a piece of music to a new therapeutic music piece that satisfies both the criteria for music therapy and the requirements of individuals. Our approach involves first developing a music feature extraction network that can extract spatio-temporal music features. We then train the network on different genres and transfer the learning results from genre classification to music transfer for anxiety reduction. We also propose a deep learning-based music transfer network with melody and accompaniment encoders and a newly designed loss function to modify a piece of music to be therapeutic. Experimental results on 92 subjects demonstrate that the transferred therapeutic music produced by the StTMTM exhibits exceptional performance in anxiety reduction. Overall, our model has the potential to make music therapy more accessible and effective for individuals with diverse backgrounds and needs.

Energy Minimization for UAV-Enabled Wireless Power Transfer and Relay Networks

In this paper, we consider an unmanned aerial vehicle (UAV)-enabled wireless power transfer (WPT) and relay communication network consisting of a base station (BS), a UAV, and multiple ground users. The UAV acts as both a wireless power transmission source and an uplink communication relay. Specifically, an entire transmission period of the considered system is divided into two stages. In the first stage, the UAV transfers the power to the ground users along a well-optimized flight trajectory and meanwhile the users transmit data to the UAV using the harvested energy. Subsequently, in the second stage, the UAV flies to the vicinity of the BS and forwards the data to the BS. For the purpose of minimizing the energy consumed by the UAV, we jointly optimize the time durations of the two stages, the UAV’s transmit powers for WPT and data forwarding, as well as its flight trajectory, subject to the constraints of the quality of service (QoS), the information forwarding, the energy causality, and the mobility of the UAV. The involved optimization problem is non-convex and highly intractable. To this end, we propose an efficient alternating algorithm to iteratively solve two subproblems with respect to the time durations of the two stages and the UAV’s transmit powers and trajectory, respectively. The first subproblem has a closed-form optimal solution and the second subproblem is handled by addressing a surrogate convex problem based on the technique of successive convex approximation. Finally, simulation results confirm the superiority of our proposed algorithm.

GMRLNet: A graph-based manifold regularization learning framework for placental insufficiency diagnosis on incomplete multimodal ultrasound data.

Multimodal analysis of placental ultrasound (US) and microflow imaging (MFI) could greatly aid in the early diagnosis and interventional treatment of placental insufficiency (PI), ensuring a normal pregnancy. Existing multimodal analysis methods have weaknesses in multimodal feature representation and modal knowledge definitions and fail on incomplete datasets with unpaired multimodal samples. To address these challenges and efficiently leverage the incomplete multimodal dataset for accurate PI diagnosis, we propose a novel graph-based manifold regularization learning (MRL) framework named GMRLNet. It takes US and MFI images as input and exploits their modality-shared and modality-specific information for optimal multimodal feature representation. Specifically, a graph convolutional-based shared and specific transfer network (GSSTN) is designed to explore intra-modal feature associations, thus decoupling each modal input into interpretable shared and specific spaces. For unimodal knowledge definitions, graph-based manifold knowledge is introduced to describe the sample-level feature representation, local inter-sample relations, and global data distribution of each modality. Then, an MRL paradigm is designed for inter-modal manifold knowledge transfer to obtain effective cross-modal feature representations. Furthermore, MRL transfers the knowledge between both paired and unpaired data for robust learning on incomplete datasets. Experiments were conducted on two clinical datasets to validate the PI classification performance and generalization of GMRLNet. State-of-the-art comparisons show the higher accuracy of GMRLNet on incomplete datasets. Our method achieves 0.913 AUC and 0.904 balanced accuracy (bACC) for paired US and MFI images, as well as 0.906 AUC and 0.888 bACC for unimodal US images, illustrating its application potential in PI CAD systems.

DualGGAN: A New Facial Style Transfer Network

In response to the background penetration problem of unsupervised style transfer algorithms in most cases, a Transformer style transfer network DualGGAN based on dual generators and fusion of relative position encoding is proposed. The network is trained using the least squares generative adversarial network, and the neural network is used as the image feature extractor to generate feature maps to obtain facial image features with attention weights from the feature maps, utilizing relative position encoding and mask loss to jointly constrain feature region style transfer. The experimental results show that the DualGGAN network effectively reduces artifact generation when implementing facial style transfer, maintains good background consistency, and has good generalization ability. Experiments have shown that the FID and KID indicators on the cat2dog and older2adult datasets are significantly improved compared to other algorithms.