Transfer Network Research Articles

A partial domain adaptation scheme based on weighted adversarial nets with improved CBAM for fault diagnosis of wind turbine gearbox

Most domain adaptation methods for fault diagnosis depend heavily on the precondition that the source and target domain have an identical label space, which is hard to be satisfied in industrial sites. Recently, many approaches have been developed to implement partial domain adaptation. However, most existing methods adopt classic convolutional neural network as the feature extractor, which limits the ability to learn discriminative representations from non-stationary vibration signals of wind turbine (WT) gearboxes. Moreover, the design of multiple subdomain adaptation will cause complex network structure with many source classes. To address these problems, this paper proposes a partial domain adaptation scheme based on weighted adversarial nets with improved convolutional block attention module (CBAM) for WT gearbox unsupervised fault diagnosis. In detail, a residual convolutional network combining the improved CBAM is designed to extract finer domain discriminative features for knowledge transfer. Meanwhile, a weighting mechanism based on the two-stage domain discriminator is designed to evaluate the contribution of each source sample, through which a simplified transfer network structure is constructed and the source samples unrelated to the target domain can be filtered. Furthermore, an adversarial transfer strategy is introduced to decrease the distribution discrepancy between domains, then the helpful diagnosis knowledge can be transferred. Experiments on two cases demonstrate the superiority and effectiveness of the proposed method compared with existing domain adaptation methods.

Domain adversarial-based multi-source deep transfer network for cross-production-line time series forecasting

Domain adversarial-based multi-source deep transfer network for cross-production-line time series forecasting

Unsupervised Vehicle Re-Identification Based on Cross-Style Semi-Supervised Pre-Training and Feature Cross-Division

Vehicle Re-Identification (Re-ID) based on Unsupervised Domain Adaptation (UDA) has shown promising performance. However, two main issues still exist: (1) existing methods that use Generative Adversarial Networks (GANs) for domain gap alleviation combine supervised learning with hard labels of the source domain, resulting in a mismatch between style transfer data and hard labels; (2) pseudo label assignment in the fine-tuning stage is solely determined by similarity measures of global features using clustering algorithms, leading to inevitable label noise in generated pseudo labels. To tackle these issues, this paper proposes an unsupervised vehicle re-identification framework based on cross-style semi-supervised pre-training and feature cross-division. The framework consists of two parts: cross-style semi-supervised pre-training (CSP) and feature cross-division (FCD) for model fine-tuning. The CSP module generates style transfer data containing source domain content and target domain style using a style transfer network, and then pre-trains the model in a semi-supervised manner using both source domain and style transfer data. A pseudo-label reassignment strategy is designed to generate soft labels assigned to the style transfer data. The FCD module obtains feature partitions through a novel interactive division to reduce the dependence of pseudo-labels on global features, and the final similarity measurement combines the results of partition features and global features. Experimental results on the VehicleID and VeRi-776 datasets show that the proposed method outperforms existing unsupervised vehicle re-identification methods. Compared with the last best method on each dataset, the method proposed in this paper improves the mAP by 0.63% and the Rank-1 by 0.73% on the three sub-datasets of VehicleID on average, and it improves mAP by 0.9% and Rank-1 by 1% on VeRi-776 dataset.

Local private actors in transnational policy networks: A relational approach to studying policy transfers in Asia

AbstractMany documented cases of policy transfer in Asia involve small and medium enterprises as well as public enterprises but little is known about how and why these local private actors participate in transnational policy transfer networks. We propose a relational view of policy transfer to examine how agency is exercised in these kinds of transfer networks. We develop a typology of the engagement of these private actors—solutionist and entrepreneur—based on their relational position in the policy transfer process and illustrate its usefulness using two cases of policy transfer between: (1) Yokohama (Japan) and Cebu (Philippines) and (2) Singapore and New Clark City (Philippines). A common theme in these networks is the diffused nature of agency where local private actors are pulled in to provide epistemic resources.

Cross-Project Transfer Learning on Lightweight Code Semantic Graphs for Defect Prediction

A deep learning system (DLS) developed based on one software project for defect prediction may well be applied to the related code on the same project but is usually difficult to be applied to new or unknown software projects. To address this problem, we propose a Transferable Graph Convolutional Neural Network (TGCNN) that can learn defects from the lightweight semantic graphs of code and transfer the learned knowledge from the source project to the target project. We discuss how the semantic graph is constructed from code; how the TGCNN can learn from the graph; and how the learned knowledge can be transferred to a new or unknown project. We also conduct a controlled experiment to evaluate our method. The result shows that despite some limitations, our method performs considerably better than existing methods.

DFTNet: Dual-Path Feature Transfer Network for Weakly Supervised Medical Image Segmentation.

Medical image segmentation has long suffered from the problem of expensive labels. Acquiring pixel-level annotations is time-consuming, labor-intensive, and relies on extensive expert knowledge. Bounding box annotations, in contrast, are relatively easy to acquire. Thus, in this paper, we explore to segment images through a novel Dual-path Feature Transfer design with only bounding box annotations. Specifically, a Target-aware Reconstructor is proposed to extract target-related features by reconstructing the pixels within the bounding box through the channel and spatial attention module. Then, a sliding Feature Fusion and Transfer Module (FFTM) fuses the extracted features from Reconstructor and transfers them to guide the Segmentor for segmentation. Finally, we present the Confidence Ranking Loss (CRLoss) which dynamically assigns weights to the loss of each pixel based on the network's confidence. CRLoss mitigates the impact of inaccurate pseudo-labels on performance. Extensive experiments demonstrate that our proposed model achieves state-of-the-art performance on the Medical Segmentation Decathlon (MSD) Brain Tumour and PROMISE12 datasets.

Improved Feature Decoupling Transfer Network Modeling based on Singular Value Decomposition for SOC Estimation in Energy-Storage Lithium-ion Batteries

Deep learning (DL) methods is applied extensively in the field of state of charge (SOC) estimation, which require training data and test data to have similar distribution. Discrepancies in data distribution arising from the complexity and diversity of lithium-ion batteries under operational conditions in practice, as well as the difficulty in obtaining data labels, make it enormously challenging to access sufficient battery data to train a specific deep estimator. Aiming to improve the performance of cross-domain SOC estimation for lithium-ion batteries, a model for SOC estimation which combines transfer learning with singular value decomposition (SVD) is proposed. To begin with, a gated recurrent unit (GRU) network is employed to avail the nonlinear dynamic characteristics of the battery from the source and target domains. Then, the features are decoupled by using SVD method to extract task-relevant, important and minor information in the network. Further, the amount of transferred information over the source network to the target network is automatically tuned by the maximum mean discrepancy (MMD) to determine the different degrees of similarity in domain, and the cosine discrepancy to measure the discrepancy on the same domain, which achieves the optimized performance of the target network.

Cross-Modality LGE-CMR Segmentation Using Image-to-Image Translation Based Data Augmentation.

Accurate segmentation of ventricle and myocardium from the late gadolinium enhancement (LGE) cardiac magnetic resonance (CMR) is an important tool for myocardial infarction (MI) analysis. However, the complex enhancement pattern of LGE-CMR and the lack of labeled samples make its automatic segmentation difficult to be implemented. In this paper, we propose an unsupervised LGE-CMR segmentation algorithm by using multiple style transfer networks for data augmentation. It adopts two different style transfer networks to perform style transfer of the easily available annotated balanced-Steady State Free Precession (bSSFP)-CMR images. Then, multiple sets of synthetic LGE-CMR images are generated by the style transfer networks and used as the training data for the improved U-Net. The entire implementation of the algorithm does not require the labeled LGE-CMR. Validation experiments demonstrate the effectiveness and advantages of the proposed algorithm.

Joint optimization of trajectory, beamforming, and power allocation in UAV-enabled WPT networks using DRL combined with water-filling algorithm

In this work, we study a wireless power transfer (WPT) network supported by an unmanned aerial vehicle (UAV), in which a UAV equipped with an array of antennas supplies wireless energy to charge ground user (GU) devices. With this configuration, we aim to maximize the lowest GU energy by jointly optimizing the UAV's trajectory, beamforming pattern, and transmitting power. Because the proposed maximization of the lowest GU energy is a non-convex optimization problem that is difficult to be solved, we transform the problem into a grid world problem in discrete-time space and propose a design of a deep reinforcement learning (DRL) approach that optimizes the loss function by determining the UAV's moving direction, beamforming angle, and transmit power level. Furthermore, we combine the water-filling algorithm with DRL, which can assist DRL in determining the hovering duration. Simulation results verify that our proposed design increases the minimum energy of GUs received from the UAV compared to the successive hover-and-fly algorithm while exploring the most appropriate path with low computational complexity.

Co-phase errors sensing method for Golay3 telescope system via a transfer network.

Optical sparse-aperture systems face severe challenges, including detecting and correcting co-phase errors. In this study, a search framework based on fine tuning a pre-trained network is proposed to analyze the co-phase errors of a Golay3 telescope system. Based on this, an error compensation control system is established. First, a hash-like binary code is created by fine-tuning the pre-trained model. Secondly, a pre-trained network is used to extract the deep features of the image, and an index database is built between the image features and the corresponding co-phase error values. Finally, the Top 1-ranked features and corresponding co-phase error values are returned through the hash-like binary code hierarchical deep search database to provide driving data for the error correction system. Numerical simulations and experimental results verify the method's validity. The experimental results show that the correction system works well when the dynamic piston is [-5,5]λ, and the tilt error range is [-15,15]µr a d. Compared with existing detection methods, this method does not require additional optical components, has a high correction accuracy, and requires a short training time. Furthermore, it can be used to detect piston and tilt errors simultaneously.

Identification of a novel lipoic acid biosynthesis pathway reveals the complex evolution of lipoate assembly in prokaryotes.

Lipoic acid is an essential biomolecule found in all domains of life and is involved in central carbon metabolism and dissimilatory sulfur oxidation. The machineries for lipoate assembly in mitochondria and chloroplasts of higher eukaryotes, as well as in the apicoplasts of some protozoa, are all of prokaryotic origin. Here, we provide experimental evidence for a novel lipoate assembly pathway in bacteria based on a sLpl(AB) lipoate:protein ligase, which attaches octanoate or lipoate to apo-proteins, and 2 radical SAM proteins, LipS1 and LipS2, which work together as lipoyl synthase and insert 2 sulfur atoms. Extensive homology searches combined with genomic context analyses allowed us to precisely distinguish between the new and established pathways and map them on the tree of life. This not only revealed a much wider distribution of lipoate biogenesis systems than expected, in particular, the novel sLpl(AB)-LipS1/S2 pathway, and indicated a highly modular nature of the enzymes involved, with unforeseen combinations, but also provided a new framework for the evolution of lipoate assembly. Our results show that dedicated machineries for both de novo lipoate biogenesis and scavenging from the environment were implemented early in evolution and that their distribution in the 2 prokaryotic domains was shaped by a complex network of horizontal gene transfers, acquisition of additional genes, fusions, and losses. Our large-scale phylogenetic analyses identify the bipartite archaeal LplAB ligase as the ancestor of the bacterial sLpl(AB) proteins, which were obtained by horizontal gene transfer. LipS1/S2 have a more complex evolutionary history with multiple of such events but probably also originated in the domain archaea.

Self-Supervised Interest Transfer Network via Prototypical Contrastive Learning for Recommendation

Cross-domain recommendation has attracted increasing attention from industry and academia recently. However, most existing methods do not exploit the interest invariance between domains, which would yield sub-optimal solutions. In this paper, we propose a cross-domain recommendation method: Self-supervised Interest Transfer Network (SITN), which can effectively transfer invariant knowledge between domains via prototypical contrastive learning. Specifically, we perform two levels of cross-domain contrastive learning: 1) instance-to-instance contrastive learning, 2) instance-to-cluster contrastive learning. Not only that, we also take into account users' multi-granularity and multi-view interests. With this paradigm, SITN can explicitly learn the invariant knowledge of interest clusters between domains and accurately capture users' intents and preferences. We conducted extensive experiments on a public dataset and a large-scale industrial dataset collected from one of the world's leading e-commerce corporations. The experimental results indicate that SITN achieves significant improvements over state-of-the-art recommendation methods. Additionally, SITN has been deployed on a micro-video recommendation platform, and the online A/B testing results further demonstrate its practical value. Supplement is available at: https://github.com/fanqieCoffee/SITN-Supplement.

Digital Platformas a Tool for the Functioning of the Technology Transfer Network

Problem setting. Today, the digital economy is being transferred to the digital platform, including in the area of technology transfer for innovative products and/or services. Analysis of recent researches and publications. The following scholars were interested in the conceptualization of digital platforms: Acs Z. J. et al. (2021) , Kenney M. and Zysman J. (2020), Plantin Jean-Ch. (2018), Van Gorp N. and Batura O. (2015), Spagnoletti P., Resca A. and Lee, G. (2015), Ghazawnehand A. and Henfridsson O. (2013), Ceccagnoli M. et al. (2012), Tiwana A. et al. (2010), Kurz T, Eder R. and Heistracher T. (2010), Venters W. (2021), Mansell R. and Steinmueller W. E. (2020), Frenken K. and Fuenfschilling L. (2020), Thompson M, Mukhopadhyay S., Bouwman H. and Jaiswal M. P. (2019), Otto B. and Jarke M. (2019), Poell T. et al. (2019), Gawer A. (2021), Van Dijck J. (2020, Keskin B. Van Dijk et al. (2018), Schwarz J. A. (2017). Among Ukrainian scholars, the problems of introducing digital platforms and the conceptual foundations of their development in the context of the formation of the digital economy were of interest to Kokhan V. P., Sichkarenko K. O., Ostrovsky I. A., Lyashenko V. I. and Vyshnevsky O. S., Semenog A. Y. Purpose of the research is to identify the existing views on the category of «digital platform» in the scientific literature, to identify its characteristics, and to provide an overview of its types. We will find out the impact of digital platforms on the activities of technology transfer networks and identify gaps in legal regulation and suggest opportunities for future research. Article’s main body. The article studies the conceptual provisions for substantiating the essence of the category «digital platform» and its impact on the activities of a technology transfer network. The author proves that a digital platform is an important factor enabling innovation through a technology transfer network, a mechanism for stimulating technology transfer, and a certain means of digital transformation. The concept of «digital platform» is widespread in the legal and economic literature, but there are significant differences of opinion among scholars regarding its essence. The study has led to the conclusion that a digital platform is a networked form of business organization, a special type of business model that brings together participants on the principles of an ecosystem and on a voluntary basis to promote valuable interaction between the involved participants, ensure coordination, dissemination of information, transfer of technology, and provision of services. The study found that there are business models of platforms that have a network effect and they exist in the following types: transactional, innovative, integrated, investment, and information platforms. The article emphasizes that from a legal perspective, a digital platform (for technology transfer) is a key tool for the functioning of a technology transfer network which provides for the connection of a large number of participants based on digital technologies, and which aims at coordinating and processing information, providing services, engaging software, assisting in finding partners for research and development and effective interaction of technology transfer participants, digitalization of business processes, transfer of new promising technologies and/or It has been established that the technology transfer network facilitates the implementation of innovative projects of national and international (transnational) technology transfer. Conclusions and prospects for the development. In order to improve the category of «digital platform», its concept should be fixed in the Law of Ukraine «On Stimulating the Development of the Digital Economy in Ukraine» of 15.07.2021 No. 1667-IX and the draft Law «On Digital Content and Digital Services», adopted as a basis on 12.01.2023. And the concept of a technology transfer network and center is defined in the Law of Ukraine «On State Regulation of Activities in the Field of Technology Transfer» of 14.09.2006 No. 143-V or in the Draft Law on Amendments to Certain Laws of Ukraine on Stimulating Activities in the Field of Technology Transfer No. 4623 of 21.01.2021.

Incorporation of Heteroatomic Fe Activates Rapid Catalytic Behaviors of Co3O4 Hollow Nanoplates Toward Advanced Lithium–Sulfur Batteries

AbstractLithium–sulfur (Li–S) battery is a highly attractive energy storage system due to its high capacity and great affordability. However, the parasitic shuttle effect and sluggish redox kinetics are perplexing the fulfillment of efficient battery electrochemistry. To tackle these challenges, herein, a hierarchical and hollow nanoplate assembled by Fe‐doped Co3O4 nanosheets (Fe–Co3O4 HHNPs) is meticulously designed as an advanced sulfur nanoreactor. The interlaced nanosheets establish a robust and porous network for fast charge transfer and efficient active site exposure. More importantly, the heteroatomic Fe incorporation tailors the electronic structure via local structure distortion and electron redistribution, contributing to massive active sites that lower the energy barrier for sulfur conversions. The resulting sulfur adsorption and catalyzation endow the Li–S cells with minimum capacity decay of 0.08% per cycle over 500 cycles and decent rate performance up to 5 C. Moreover, a high areal capacity of 9.0 mAh cm−2 after 55 cycles is also achievable under raised sulfur loading of 11 mg cm−2 and limited electrolyte (E/S = 3.9 µL mg−1). This work provides an elaborate and instructive paradigm for designing catalytic nanoreactors toward superior Li–S electrochemistry.

SRPSGAN: Super-resolution with pose and expression robust spatial-aware generative adversarial network for makeup transfer

SRPSGAN: Super-resolution with pose and expression robust spatial-aware generative adversarial network for makeup transfer

A universal transfer network for machinery fault diagnosis

Domain adaptation (DA) methods have achieved promising results in machinery fault diagnosis owing to their ability to mitigate the distribution discrepancy between domains. However, existing fault diagnosis methods based on DA are tailored for a specific setting, and highly rely on prior knowledge about the relationship between the source and target label sets which is usually not available in advance. To broaden the applicability of DA for fault diagnosis, this paper proposes a universal transfer network to handle all types of DA settings, including closed-set DA, partial DA, open-set DA, and open-partial DA. The proposed method utilizes self-supervised learning to uncover the cluster structure of the target domain, and incorporates entropy-based feature alignment to align shared-class samples while separating unknown-class samples. Moreover, an open-set classifier is trained to provide a confidence criterion, which is then used to construct a sample-level uncertainty criterion for identifying unknown-class samples efficiently. The proposed method is evaluated on Office-31 dataset and two fault diagnosis datasets. Our experimental results demonstrate that the proposed method performs better in all DA settings when compared to other methods.

Rational construction of rich coordination-unsaturated Zr-BTB electrocatalyst towards advanced lithium-sulfur batteries

Lithium-sulfur (Li-S) batteries show great potential to achieve high-density energy storage, but their commercial application is severely hindered by the notorious shuttling effect of lithium polysulfides (LiPS) and their sluggish conversion kinetics. Herein, we develop a Quasi Zr-based metal–organic frameworks with the tritopic carboxylate ligand 1,3,5-tris(4-carboxyphenyl)benzene (BTB) microflowers with cross-linked CNTs (Q-Zr-BTB@CNTs). The deficiency of Zr-O coordination at the Zr6 node is realized via a control of the deligandation behavior, which subtly modulates the local environment and electronic structure towards enhanced chemical affinity to polysulfides. Meanwhile, the interweave of the ultrathin Q-Zr-BTB nanosheets and CNTs establishes a porous and conductive network for fast charge transfer as well as a vast exposure of active interfaces. As a result, such a combination of defect engineering and architecture construction imposes strong sulfur adsorption and catalyzation, endowing the S/Q-Zr-BTB@CNTs cathode with a high initial capacity (1130 mAh g−1 at 0.2 C) and an excellent cycling stability (500 cycles at 1 C with 0.05% decay per cycle), as well as a decent electrochemical performance even under a high sulfur loading up to 12.7 mg cm−2. This study provides a novel strategy of designing unsaturated coordination centers to boost sulfur catalysis for high-performance lithium sulfur batteries.

Deep Learning-Enhanced Hybrid Fruit Fly Optimization for Intelligent Traffic Control in Smart Urban Communities

The rapid urbanization accompanying the evolution into “smart” communities presents numerous challenges, not least of which is the significant increase in road vehicles. This proliferation exacerbates congestion and accident rates, posing major barriers to the successful implementation of innovative technologies such as wireless sensor networks (WSNs), surveillance cameras, and the Internet of Things (IoT). Accurate traffic flow prediction, a crucial component of these technological initiatives, requires a reliable and efficient methodology.This research explores the implementation of an intelligent traffic control system that employs a Transferable Texture Convolutional Neural Network (TTCNN). The design of this system eschews the traditional pooling layer, instead incorporating three convolutional layers and a single energy layer (EL). This configuration facilitates the provision of real-time traffic updates, which can enhance the utility and efficiency of the smart city infrastructure.A model inspired by the hybrid fruit fly (HFFO) optimizes the system's hyperparameters. The application of HFFO to the TTCNN showcases the potential for improved accuracy in traffic flow prediction. Simulation results suggest that the HFFO provides superior organizational boundaries for the TTCNN, enhancing the overall accuracy of the model's predictions. The hybrid forecasting method discussed herein demonstrates its potential to outperform other established techniques.This investigation sheds light on the potential benefits of applying deep learning algorithms and hybrid models in the context of traffic flow prediction and control, contributing to the ongoing development of smart urban communities.

Health and lifespan prediction considering degradation patterns of lithium-ion batteries based on transferable attention neural network

With the continuous concern on the safety of battery systems, accurate and rapid assessment of battery degradation is essential for practical applications. In this paper, a transferable attention network model based on deep learning is developed to evaluate battery degradation, which can simultaneously predict state of health (SOH) and remaining useful life (RUL) for lithium-ion batteries. First, degradation patterns of the cells are identified by K-means clustering based on the difference of the cells at their early cycles. Secondly, the attention mechanisms are designed to suppress noises in extracted feature maps and trace the interaction between long- and short-term degradation modes. Thirdly, the common knowledge represented by the reference cells and the unique degradation features of the target cell are fused by transfer learning, then SOH and RUL prediction are realized through multi-task learning. Finally, a large-scale battery dataset containing different cycle conditions is used to verify the accuracy and generalization of the developed method. The results show that the developed method achieves accurate SOH and RUL prediction with the average root mean square error within 0.14% and six cycles, respectively.

Cross-city Landuse classification of remote sensing images via deep transfer learning

This paper describes a deep transfer model which consists of multiple sub-networks which are independently optimized by a supervised task-oriented loss and an unsupervised consistency loss. The former loss function utilizes annotations to accomplish the designed goal. The latter loss prompts the network to learn the target domain data distribution and encourages multiple sub-networks to share learned knowledge. We utilize the proposed model to work with a global local climate zone classification. For the dataset, the source domain includes 352,366 training samples from 42 cities, and the target domain has 48,307 samples from 10 other cities. According to our experiments, the proposed model improves 3.46% overall accuracy and 2.97% average accuracy when compared with other state-of-the-art domain adaptation methodologies. Besides, the classification maps also visualize the outstanding performance of the proposed deep transfer network.