Intermediate Domains Research Articles

Shh signaling directs dorsal ventral patterning in the regenerating X. tropicalis spinal cord.

Tissue development and regeneration rely on the deployment of embryonic signals to drive progenitor activity and thus generate complex cell diversity and organization. One such signal is Sonic Hedgehog (Shh), which establishes the dorsal-ventral (D/V) axis of the spinal cord during embryogenesis. However, the existence of this D/V axis and its dependence on Shh signaling during regeneration varies by species. Here we investigate the function of Shh signaling in patterning the D/V axis during spinal cord regeneration in Xenopus tropicalis tadpoles. We find that neural progenitor markers Msx1/2, Nkx6.1, and Nkx2.2 are confined to dorsal, intermediate and ventral spatial domains, respectively, in both the uninjured and regenerating spinal cord. These domains are altered by perturbation of Shh signaling. Additionally, we find that these D/V domains are more sensitive to Shh perturbation during regeneration than uninjured tissue. The renewed sensitivity of these neural progenitor cells to Shh signals represents a regeneration specific response and raises questions about how responsiveness to developmental patterning cues is regulated in mature and regenerating tissues.

Bridging the Source-to-Target Gap for Cross-Domain Person Re-identification with Intermediate Domains

Bridging the Source-to-Target Gap for Cross-Domain Person Re-identification with Intermediate Domains

Transfer learning via intermediate structures

Recent advances have been made in the field of population-based structural health monitoring (PBSHM), which seeks to share valuable information across a population to improve inferences regarding the health states of the members. However, enabling knowledge transfer between structures with highly disparate features (i.e., heterogeneous populations) is an ongoing challenge. The current work proposes a technique for knowledge transfer within a heterogeneous population, via intermediate structures that help bridge the gap in information between the structures of interest. The transformation from the source to the intermediate and target domains is based on geodesic flows, to incrementally map between changes in geometric and statistical properties. The proposed technique is demonstrated using simulated lumped-mass systems.

Bearing fault diagnosis using Gradual Conditional Domain Adversarial Network

Given the limited availability of accurately labeled data in fault diagnosis across various industrial scenarios, we proposed a Gradual Conditional Domain Adversarial Network (GCDAN) incorporating various fault categories and rotating speeds. We constructed a prototype system for collecting three-dimensional vibration data samples and modified the network structure to accommodate the input. Inspired by the generalization capability of cross-device scenarios, we adopted CDAN as the main component. To overcome the performance degradation caused by the source and target domains with substantial distribution differences, we introduced Gradual Domain Adaptation into our algorithm. Unlabeled data samples obtained from the intermediate domains were used to train a sequence of CDANs. Experimental comparison results confirmed the effectiveness of the 3-D input data and its network alteration. Additionally, GCDAN performed better over challenging transfer tasks compared to the existing state-of-art algorithms in terms of prediction accuracy and multi-class classification metrics.

Reconstitution of cytolinker-mediated crosstalk between actin and vimentin

Cell shape and motility are determined by the cytoskeleton, an interpenetrating network of actin filaments, microtubules, and intermediate filaments. The biophysical properties of each filament type individually have been studied extensively by cell-free reconstitution. By contrast, the interactions between the three cytoskeletal networks are relatively unexplored. They are coupled via crosslinkers of the plakin family such as plectin. These are challenging proteins for reconstitution because of their giant size and multidomain structure. Here we engineer a recombinant actin-vimentin crosslinker protein called ‘ACTIF’ that provides a minimal model system for plectin, recapitulating its modular design with actin-binding and intermediate filament-binding domains separated by a coiled-coil linker for dimerisation. We show by fluorescence and electron microscopy that ACTIF has a high binding affinity for vimentin and actin and creates mixed actin-vimentin bundles. Rheology measurements show that ACTIF-mediated crosslinking strongly stiffens actin-vimentin composites. Finally, we demonstrate the modularity of this approach by creating an ACTIF variant with the intermediate filament binding domain of Adenomatous Polyposis Coli. Our protein engineering approach provides a new cell-free system for the biophysical characterization of intermediate filament-binding crosslinkers and for understanding the mechanical synergy between actin and vimentin in mesenchymal cells.

Inter-Domain Invariant Cross-Domain Object Detection Using Style and Content Disentanglement for In-Vehicle Images

The accurate detection of relevant vehicles, pedestrians, and other targets on the road plays a crucial role in ensuring the safety of autonomous driving. In recent years, object detectors based on Transformers or CNNs have achieved excellent performance in the fully supervised paradigm. However, when the trained model is directly applied to unfamiliar scenes where the training data and testing data have different distributions statistically, the model’s performance may decrease dramatically. To address this issue, unsupervised domain adaptive object detection methods have been proposed. However, these methods often exhibit decreasing performance when the gap between the source and target domains increases. Previous works mainly focused on utilizing the style gap to reduce the domain gap while ignoring the content gap. To tackle this challenge, we introduce a novel method called IDI-SCD that effectively addresses both the style and content gaps simultaneously. Firstly, the domain gap is reduced by disentangling it into the style gap and content gap, generating corresponding intermediate domains in the meanwhile. Secondly, during training, we focus on one single domain gap at a time to achieve inter-domain invariance. That is, the content gap is tackled while maintaining the style gap, and vice versa. In addition, the style-invariant loss is used to narrow down the style gap, and the mean teacher self-training framework is used to narrow down the content gap. Finally, we introduce a multiscale fusion strategy to enhance the quality of pseudo-labels, which mainly focus on enhancing the detection performance for extreme-scale objects (very large or very small objects). We conduct extensive experiments on four mainstream datasets of in-vehicle images. The experimental results demonstrate the effectiveness of our method and its superiority over most of the existing methods.

FDGNet: Frequency Disentanglement and Data Geometry for Domain Generalization in Cross-Scene Hyperspectral Image Classification.

Cross-scene hyperspectral image classification (HSIC) poses a significant challenge in recognizing hyperspectral images (HSIs) from different domains. The current mainstream approaches based on domain adaptation (DA) methods need to access target data when aligning distributions between domains, limiting the applicability of the model. In contrast, recent domain generalization (DG) methods aim to directly generalize to unseen domains, eliminating the requirements for target data during training. Nonetheless, most DG-based methods overly focus on randomizing sample styles, leading to semantically compromised samples. In addition, broadening the source distribution without ensuring reasonable support may result in undesired extended distributions. To address these issues, we propose a novel DG network with frequency disentanglement and data geometry (FDGNet) for cross-scene HSIC. Specifically, we first develop a spectral-spatial encoder based on frequency disentanglement (FDSS encoder), which facilitates synthesized domains to preserve their semantic consistency while simulating interdomain gaps with the source domain. Second, to avoid the generation of unrealistic samples, we incorporate data geometry into adversarial training. This helps diversify new domains while keeping the data geometry of extended domains in an explainable support. To improve the learning of domain-invariant representation, we propose an intermediate domain sampling strategy based on the class-wise perceptual manifold. This strategy synthesizes reliable intermediate domains by sampling from class-wise manifold flows estimated over the source and extended domains. Extensive experiments and analysis on three public HSI datasets yield the superiority of our proposed FDGNet. The codes will be available from the website: https://github.com/Qba-heu/FDGNet.

Developing a spectrum model of engagement in services for first episode psychosis: beyond attendance.

Early intervention services (EIS) for psychosis have proven highly effective in treating first episode psychosis. Yet, retention or "engagement" in EIS remains highly variable. Dis/engagement as a contested concept and fluid process involving relationships between service providers and service users remains poorly understood. This study aimed to critically evaluate and explain the dynamic interplay of service provider-user relationships in effecting dis/engagement from an early intervention program for psychosis. Forty study participants, 16 service providers and 24 service users (19 current and 5 disengaged) from a Canadian EIS program, were administered semi-structured interviews. Qualitative analysis was conducted using grounded theory methods, with findings captured and reconceptualized in a novel explanatory model. A model of engagement with eight major domains of engagement in EIS positioned along a control-autonomy spectrum was developed from the findings, with Clinical engagement (attendance) and Life engagement (life activities) at opposite ends of the spectrum, interspersed by six intermediate domains: Medication/treatment, Symptoms/illness, Mental health, Physical health/wellness, Communication, and Relationships, each domain bearing uniquely on engagement. An examination of service user and service provider perspectives on the various domains identified in the spectrum model, and their dynamic interplay, reveals the complexity of choices faced by service users in engaging and not engaging with services.

Discriminability-Aware Intermediate Domains for Mismatched Steganalysis

Discriminability-Aware Intermediate Domains for Mismatched Steganalysis

Universal Mismatched Steganalysis Equipped With Progressive Intermediate Domains

Universal Mismatched Steganalysis Equipped With Progressive Intermediate Domains

The Geometry of Low- and High-Level Perceptual Spaces.

Low-level features are typically continuous (e.g., the gamut between two colors), but semantic information is often categorical (there is no corresponding gradient between dog and turtle) and hierarchical (animals live in land, water, or air). To determine the impact of these differences on cognitive representations, we characterized the geometry of perceptual spaces of five domains: a domain dominated by semantic information (animal names presented as words), a domain dominated by low-level features (colored textures), and three intermediate domains (animal images, lightly texturized animal images that were easy to recognize, and heavily texturized animal images that were difficult to recognize). Each domain had 37 stimuli derived from the same animal names. From 13 participants (9F), we gathered similarity judgments in each domain via an efficient psychophysical ranking paradigm. We then built geometric models of each domain for each participant, in which distances between stimuli accounted for participants' similarity judgments and intrinsic uncertainty. Remarkably, the five domains had similar global properties: each required 5-7 dimensions, and a modest amount of spherical curvature provided the best fit. However, the arrangement of the stimuli within these embeddings depended on the level of semantic information: dendrograms derived from semantic domains (word, image, and lightly texturized images) were more "tree-like" than those from feature-dominated domains (heavily texturized images and textures). Thus, the perceptual spaces of domains along this feature-dominated to semantic-dominated gradient shift to a tree-like organization when semantic information dominates, while retaining a similar global geometry.

Influence of amorphous intermediate domains on the performance of Na2O–Bi2O3–GeO2 glass–ceramic anode material system

Influence of amorphous intermediate domains on the performance of Na2O–Bi2O3–GeO2 glass–ceramic anode material system

A novel cross-bearing fault diagnosis method based on pseudo-label transitive domain adaptation networks

Transfer across bearings produces a greater domain shift than transfer across working conditions (WCs). Because different bearings may have differences in structural parameters, measurement environments, and WCs, a direct transfer may significantly lower the diagnostic accuracy of the target bearing. A novel fault diagnosis method based on pseudo-label transitive domain adaptation networks (PLTDANs) is proposed to address this problem. First, empirical selection criteria are offered to ensure that appropriate intermediate domains are selected and can better bridge the source and target domains. The intermediate-domain-added TDAN can split the direct transfer process into source-intermediate and intermediate-target transfer diagnoses. This division allows for the gradual correction of the domain shift. Second, the cross-domain pseudo-label constraint (CDPLC) is proposed to select high-confidence intermediate domain samples and generate corresponding pseudo-labels. Pseudo-labels highlight the health status of intermediate domain samples. The application of CDPLC aims to minimize the cumulative error of the TDAN. A cross-bearing fault diagnosis experiment demonstrated the PLTDAN's effectiveness.

Sodium-Bismuth-Titanium glass–ceramic network: A high capacity anode network for Na+ ion storage

This article discusses 95[35Na2O-65Bi2O3]: 5TiO2 (mol%) glass-ceramic anode material system (abbreviated as GC-NBT5.0) in order to overcome major issues with Na-ion battery technology, such as inadequate reversible capacity, rate capability, electrical conductivity, and safety. After heat treatment for different heat treating regimes, the production of intense intermediate domains Na3Bi and Na3Ti results in flexible active centers with which heavier Na+ ions may react very easily for Na+ ion storage in the glass-ceramic anode network. As evaluated at 50mA/g over a longer cycle life of up to 3000 cycles, the GC-NBT5.0-8h glass ceramic anode beats both glass anode and traditional graphite anodes with a specific discharge capacity of 801mAh/g. However, the cell design is still backed by the best electrical conductivity because of the good interfacial contact between the glass-ceramic anode and electrolyte, and the lowest cycle-dependent complex impedance characteristics (Rct and Zw), which guarantee consistent capacity retention and the highest Columbic efficiency over longer cycle durations.

Cost-effective framework for gradual domain adaptation with multifidelity

In domain adaptation, when there is a large distance between the source and target domains, the prediction performance will degrade. Gradual domain adaptation is one of the solutions to such an issue, assuming that we have access to intermediate domains, which shift gradually from the source to the target domain. In previous works, it was assumed that the number of samples in the intermediate domains was sufficiently large; hence, self-training was possible without the need for labeled data. If the number of accessible intermediate domains is restricted, the distances between domains become large, and self-training will fail. Practically, the cost of samples in intermediate domains will vary, and it is natural to consider that the closer an intermediate domain is to the target domain, the higher the cost of obtaining samples from the intermediate domain is. To solve the trade-off between cost and accuracy, we propose a framework that combines multifidelity and active domain adaptation. The effectiveness of the proposed method is evaluated by experiments with real-world datasets.

Multiround Transfer Learning and Modified Generative Adversarial Network for Lung Cancer Detection

Lung cancer has been the leading cause of cancer death for many decades. With the advent of artificial intelligence, various machine learning models have been proposed for lung cancer detection (LCD). Typically, challenges in building an accurate LCD model are the small-scale datasets, the poor generalizability to detect unseen data, and the selection of useful source domains and prioritization of multiple source domains for transfer learning. In this paper, a multiround transfer learning and modified generative adversarial network (MTL-MGAN) algorithm is proposed for LCD. The MTL transfers the knowledge between the prioritized source domains and target domain to get rid of exhaust search of datasets prioritization among multiple datasets, maximizing the transferability with a multiround transfer learning process, and avoiding negative transfer via customization of loss functions in the aspects of domain, instance, and feature. In regard to the MGAN, it not only generates additional training data but also creates intermediate domains to bridge the gap between the source domains and target domains. 10 benchmark datasets are chosen for the performance evaluation and analysis of the MTL-MGAN. The proposed algorithm has significantly improved the accuracy compared with related works. To examine the contributions of the individual components of the MTL-MGAN, ablation studies are conducted to confirm the effectiveness of the prioritization algorithm, the MTL, the negative transfer avoidance via loss functions, and the MGAN. The research implications are to confirm the feasibility of multiround transfer learning to enhance the optimal solution of the target model and to provide a generic approach to bridge the gap between the source domain and target domain using MGAN.

Multi-Source Domain Fusion Cross-Domain Pedestrian Recognition Based on High-Quality Intermediate Domains

Multi-Source Domain Fusion Cross-Domain Pedestrian Recognition Based on High-Quality Intermediate Domains

Asynchronous Generative Adversarial Network for Asymmetric Unpaired Image-to-Image Translation

The unpaired image-to-image translation aims to translate input images from one source domain to some desired outputs in a target domain by learning from unpaired training data. Cycle-consistency constraint provides a general principle to estimate and measure forward and backward mapping functions between two domains. In many cases, the information entropy of images from the two domains is not equal, resulting in an information-rich domain and an information-poor domain. However, existing solutions based on cycle-consistency either completely discard the information asymmetry between the two domains (a common choice), which leads to inferior translation performance for the asymmetric unpaired image-to-image translation, or have to rely on special task-specific designs and introduce extra loss components. These elaborative designs especially for the relatively harder translation direction from the information-poor domain to the information-rich domain (poor-to-rich translation) require extra labor and are limited to some specific tasks. In this paper, we propose a novel asynchronous generative adversarial network named Async-GAN, which provides a model-agnostic framework for easily turning symmetrical models into powerful asymmetric counterparts that can handle asymmetric unpaired image-to-image translation much better. The key innovation is to iteratively build gradually improving intermediate domains for generating pseudo paired training samples, which provide stronger full supervision for assisting the poor-to-rich translation. Extensive experiments on various asymmetric unpaired translation tasks demonstrate the superiority of the proposal. Furthermore, the proposed training framework could be extended to various Cycle-GAN solutions and achieve a performance gain.

Factor XII Explored with AlphaFold - Opportunities for Selective Drug Development.

Medical device associated thrombosis is an important clinical problem. This type of thrombosis can result from Factor XII (FXII) binding to non-natural surface materials and subsequent activation of the contact pathway. This drives the development of new therapeutic strategies to block this pathway and information on the structural properties of FXII should catalyse this quest. Presently, there is no publicly available crystal structure of full-length FXII. However, the AlphaFold Protein Structure Database provides a model structure. We here explore this model in combination with previous structure-function studies to identify opportunities for selective pharmacological blockade of the contribution of FXII in medical device associated thrombosis. Previous studies demonstrated that FXII activation is dependent on molecular cleavage after R353. We subsequently proposed that protein conformation protects this cleavage site to ensure zymogen quiescence and prevent inappropriate FXII activation. The AlphaFold model shows that a small loop containing R353 indeed is buried in the globular molecule. This is the result of intra-molecular interactions between the (N-terminal) Fibronectin type II domain, (central) kringle and (C-terminal) protease domain, in a structure that resembles a three-point harness. Furthermore, this interaction pushes the intermediate domains, as well as the flexible proline-rich region (PRR), outward while encapsulating R353 in the molecule. The outward directed positively charged patches are likely to be involved in binding to anionic surfaces. The binding of FXII to surfaces (and several monoclonal antibodies) acccelerates its activation by inducing conformational changes. For prevention of medical device associated thrombosis, it is therefore important to target the surface binding sites of FXII without causing structural changes.

CNNs/ViTs-CNNs/ViTs: Mutual distillation for unsupervised domain adaptation

Unsupervised Domain Adaptation (UDA) is a popular machine learning technique to reduce the distribution discrepancy among domains. In previous UDA methods, only convolutional neural networks (CNNs) or vision transformers (ViTs) are used as the backbone. Therefore, these methods are subjected to inherent characteristics of single-CNNs or single-ViTs. In fact, CNNs are more biased towards local textures, while ViTs are inclined to learn the shape information of images. In this paper, we consider that the model combining the above two bias properties is closer to the human visual neural system and achieves more robust performance. A novel mutual distillation method for UDA without any assistant networks is proposed. Mutual distillation between two backbones with complementary properties (i.e. CNNs & ViTs) can promote each other, leading to better domain knowledge transfer. Additionally, traditional domain-mixup approaches can only mix limited cross-domain information through linear interpolation. To encourage more cross-domain information interaction between the two backbones and bridge the domain gap, we propose a patch-mixup method, building a train of mixed intermediate domains composed of augmented patches. Meanwhile, a cross-domain semantic alignment loss is proposed to align the semantic information of these domains. Extensive experiments show that our method achieves state-of-the-art results on several standard UDA benchmarks, such as Office-31, Office-Home, ImageCLEF-DA and VisDA-2017.