Non-local Dependency Research Articles

A multi-view graph contrastive learning framework for deciphering spatially resolved transcriptomics data.

Spatially resolved transcriptomics data are being used in a revolutionary way to decipher the spatial pattern of gene expression and the spatial architecture of cell types. Much work has been done to exploit the genomic spatial architectures of cells. Such work is based on the common assumption that gene expression profiles of spatially adjacent spots are more similar than those of more distant spots. However, related work might not consider the nonlocal spatial co-expression dependency, which can better characterize the tissue architectures. Therefore, we propose MuCoST, a Multi-view graph Contrastive learning framework for deciphering complex Spatially resolved Transcriptomic architectures with dual scale structural dependency. To achieve this, we employ spot dependency augmentation by fusing gene expression correlation and spatial location proximity, thereby enabling MuCoST to model both nonlocal spatial co-expression dependency and spatially adjacent dependency. We benchmark MuCoST on four datasets, and we compare it with other state-of-the-art spatial domain identification methods. We demonstrate that MuCoST achieves the highest accuracy on spatial domain identification from various datasets. In particular, MuCoST accurately deciphers subtle biological textures and elaborates the variation of spatially functional patterns.

Contrastive Learning-Based Dual Dynamic GCN for SAR Image Scene Classification.

As a typical label-limited task, it is significant and valuable to explore networks that enable to utilize labeled and unlabeled samples simultaneously for synthetic aperture radar (SAR) image scene classification. Graph convolutional network (GCN) is a powerful semisupervised learning paradigm that helps to capture the topological relationships of scenes in SAR images. While the performance is not satisfactory when existing GCNs are directly used for SAR image scene classification with limited labels, because few methods to characterize the nodes and edges for SAR images. To tackle these issues, we propose a contrastive learning-based dual dynamic GCN (DDGCN) for SAR image scene classification. Specifically, we design a novel contrastive loss to capture the structures of views and scenes, and develop a clustering-based contrastive self-supervised learning model for mapping SAR images from pixel space to high-level embedding space, which facilitates the subsequent node representation and message passing in GCNs. Afterward, we propose a multiple features and parameter sharing dual network framework called DDGCN. One network is a dynamic GCN to keep the local consistency and nonlocal dependency of the same scene with the help of a node attention module and a dynamic correlation matrix learning algorithm. The other is a multiscale and multidirectional fully connected network (FCN) to enlarge the discrepancies between different scenes. Finally, the features obtained by the two branches are fused for classification. A series of experiments on synthetic and real SAR images demonstrate that the proposed method achieves consistently better classification performance than the existing methods.

Processing covert dependency: An eye-tracking study of scope interpretations of embedded Wh-questions in Mandarin.

Non-local dependency in Mandarin wh-questions has been extensively researched in theoretical linguistics, but it remains an under-studied topic in the field of language processing. Unlike languages that require wh-movement to form wh-questions, Mandarin is a wh-in-situ language, and hence is generally assumed to require a covert dependency between a wh-phrase and its scope-bearing position. Mandarin therefore provides an ideal linguistic environment in which to study not only cognitive-processing mechanisms, but also how different types of non-local dependency, especially covert dependency, can be handled by readers. This paper investigates the processing of such covert non-local dependency in multiple embedded clauses, that is, multiple complementizer phrases (CPs). In wh-in-situ sentences with multiple CPs, the wh-phrases' scope varies according to the types of verbs and their embedded clauses. Based on the subcategorization of clausal verbs, we designed four experimental conditions: double-embedded low scope, double-embedded high scope, double-embedded ambiguous scope, and long distance in pivotal construction. According to memory-based and distance-based language processing theories, the low-scope condition should be easier to process than the high-scope one, because the former has a shorter linear distance than the latter when forming dependencies; and pivotal construction should be easier to process than high-scope embedded clauses, because the former has a shorter structural distance. In cases where both low- and high-scope interpretations are possible, we aim to determine whether readers exhaust every potential interpretation during comprehension, or adopt a 'good-enough' approach to obtaining an interpretation via an easier and less costly process. To this end, we will adopt the eye-tracking technique that allows us to obtain fine-grained reading-time data, which can be used to compare processing across conditions. The results will contribute to understanding human readers' mechanisms for processing covert dependency and resolving scope ambiguity in wh-in-situ languages.

TMS-GAN: A Twofold Multi-Scale Generative Adversarial Network for Single Image Dehazing

In recent years, learning-based single image dehazing networks have been comprehensively developed. However, performance improvement is limited due to domain shift between trained synthetic hazy images and untrained real-world hazy images. To alleviate this issue, this paper proposes a real-world dehazing targeted training scheme which nearly realizes paired real-world data training. As a result, a Twofold Multi-scale Generative Adversarial Network (TMS-GAN) consisting of a Haze-generation GAN (HgGAN) and a Haze-removal GAN (HrGAN) is designed. HgGAN attributes real haze properties to synthetic images and HrGAN removes haze from both synthetic and generated fake realistic data under supervision. Thus, the proposed method can better adapt to real-world image dehazing using this cooperative training scheme. Meanwhile, several structural advances of TMS-GAN also improve dehazing performance. Specifically, a haze residual map based on atmospheric scattering model is deduced in HgGAN for fake realistic data generation. The dual-branch generator in HrGAN draws attention to detail restoration by one branch along with another color-branch. A plug-and-play Multi-attention Progressive Fusion Module (MAPFM) is proposed and inserted in both HgGAN and HrGAN. MAPFM incorporates multi-attention mechanism to guide multi-scale feature fusion in a progressive manner, in which Adjacency-attention Block (AAB) can capture contributing features of each level and Self-attention Block (SAB) can establish non-local dependency of feature fusion. Experiments on mainstream benchmarks show that the proposed framework is superior especially on real-world hazy images among single image dehazing methods.

Eighteen-month-old infants represent nonlocal syntactic dependencies

Significance As mature speakers of a language, we are able to produce and understand an indefinite number of sentences. This ability comes from a powerful cognitive system—syntax—whose properties reveal the types of computations that human minds can engage in. One core property is the capacity to encode abstract grammatical dependencies that can hold at a distance. When does this property emerge in development? We identify a nonlocal dependency that 18-mo-old infants, but not younger infants, represent syntactically. These findings suggest that the second year of life is a period of active syntactic development and that even before they regularly produce full sentences of their own, infants demonstrate some of the core computational capacities that syntax relies on.

Document-level relation extraction via graph transformer networks and temporal convolutional networks

• We propose a novel document-level relation extraction framework DocRE-HGNN. • Temporal convolutional networks is used to encode document with non-local dependency. • Heterogeneous GNN is adopted to obtain entity-entity representation automatically. • DocRE-HGNN can obtain both long inter-sentence paths and short intra-sentence paths. • Experiments on document-level biomedical datasets demonstrate model’s effectiveness. Relation Extraction (RE) aims at extracting meaningful relation facts between entities in texts. It is an important semantic processing task in the field of natural language processing (NLP) and has many applications. Traditional RE focuses on extracting entity relationships from a single input sentence. Recently, the research scope has been extended from sentence level to document level. However, compared with sentence-level RE, document-level RE, which needs to identify the inter-sentence relations from entities scattered in different sentences, is more complex and still lacks of solutions. To solve this problem, we propose a novel document-level RE method based on Heterogeneous Graph Neural Networks in this paper. Concretely, to obtain token embeddings containing long-distance dependency signals well, we encode the document with Temporal Convolutional Networks, whose dilated convolution and residual structure allow the effective and efficient preservation of historical information. To better describe the interaction between different elements, we construct the input documents as heterogeneous graphs with different node and edge types and utilize Graph Transformer Networks to generate semantic paths. Numerical experiments on two document-level biomedical datasets demonstrate the effectiveness of the proposed method.

Local and non-local dependency learning and emergence of rule-like representations in speech data by deep convolutional generative adversarial networks

This paper argues that training Generative Adversarial Networks (GANs) on local and non-local dependencies in speech data offers insights into how deep neural networks discretize continuous data and how symbolic-like rule-based morphophonological processes emerge in a deep convolutional architecture. Acquisition of speech has recently been modeled as a dependency between latent space and data generated by GANs in Beguš (2020b), who models learning of a simple local allophonic distribution. We extend this approach to test learning of local and non-local phonological processes that include approximations of morphological processes. We further parallel outputs of the model to results of a behavioral experiment where human subjects are trained on the data used for training the GAN network. Four main conclusions emerge: (i) the networks provide useful information for computational models of speech acquisition even if trained on a comparatively small dataset of an artificial grammar learning experiment; (ii) local processes are easier to learn than non-local processes, which matches both behavioral data in human subjects and typology in the world’s languages. This paper also proposes (iii) how we can actively observe the network’s progress in learning and explore the effect of training steps on learning representations by keeping latent space constant across different training steps. Finally, this paper shows that (iv) the network learns to encode the presence of a prefix with a single latent variable; by interpolating this variable, we can actively observe the operation of a non-local phonological process. The proposed technique for retrieving learning representations has general implications for our understanding of how GANs discretize continuous speech data and suggests that rule-like generalizations in the training data are represented as an interaction between variables in the network’s latent space.

Nonlinear vibration analysis of two-phase local/nonlocal nanobeams with size-dependent nonlinearity by using Galerkin method

It has been proved that using pure nonlocal elasticity, especially in differential form, leads to inconsistent and unreliable results. Therefore, to obviate these weaknesses, Eringen’s two-phase local/nonlocal elasticity has been recently used by researchers to consider the nonlocal size dependency of nanostructures. Given this, for the first time, the size-dependent nonlinear free vibration of nanobeams is investigated in this article within the framework of two-phase elasticity by using the Galerkin method. Contrary to differential nonlocal elasticity, the size dependency of the axial tension force, due to von Karman nonlinearity, is considered, and its effect on the nonlinear vibration is examined. The correct procedure of using the Galerkin method for studying the nonlinear vibration of two-phase nanobeams is introduced. It is shown that, although it is possible to extract the linear mode shapes of two-phase nanobeams using its equal differential equation, integral form of two-phase elasticity should be considered in the Galerkin method. Furthermore, it is observed that in two-phase elasticity, applying classic mode shapes in the Galerkin method leads to significant errors, especially in higher nonlocal parameters and lower values of local phase fraction and amplitude ratios.

Nonlocal regularized CNN for image segmentation

Non-local dependency is a very important prior for many image segmentation tasks. Generally, convolutional operations are building blocks that process one local neighborhood at a time which means the convolutional neural networks(CNNs) usually do not explicitly make use of the non-local prior on image segmentation tasks. Though the pooling and dilated convolution techniques can enlarge the receptive field to use some nonlocal information during the feature extracting step, there is no nonlocal priori for feature classification step in the current CNNs' architectures. In this paper, we present a non-local total variation (TV) regularized softmax activation function method for semantic image segmentation tasks. The proposed method can be integrated into the architecture of CNNs. To handle the difficulty of back-propagation for CNNs due to the non-smoothness of nonlocal TV, we develop a primal-dual hybrid gradient method to realize the back-propagation of nonlocal TV in CNNs. Experimental evaluations of the non-local TV regularized softmax layer on a series of image segmentation datasets showcase its good performance. Many CNNs can benefit from our proposed method on image segmentation tasks.

Nonlocal strain gradient nonlinear resonance of bi-directional functionally graded composite micro/nano-beams under periodic soft excitation

With the aid of advanced design techniques, functionally graded materials as promising new materials can be fabricated into various micro/nano-structures to acquire stronger mechanical performance. In this work, the size-dependent nonlinear primary resonance of periodic soft excited micro/nano-beams made of bi-directional functionally graded materials (2D-FGMs) is studied. To accomplish this end, the nonlocal strain gradient theory of elasticity is utilized within the framework of the refined hyperbolic shear deformation beam theory to construct a size-dependent beam model. On the basis of the variational approach using the principle of Hamilton, the non-classical differential equations of motion are achieved. Thereafter, a discretization scheme based numerical solving process via generalized differential quadrature method (GDQM) together with the pseudo-arclength continuation technique, the nonlocal strain gradient frequency response and amplitude response associated with the nonlinear primary resonance of 2D-FGM micro/nano-beams with different boundary conditions are obtained. It is displayed that the nonlocal size dependency makes a reduction in the oscillation amplitudes associated with both of the bifurcation points, but the strain gradient size effect causes to increase them. These patterns are more significant for the second bifurcation point and for the simply supported-simply supported boundary conditions in comparison with the clamped-clamped ones. Also, it is observed that by increasing the value of both of the axial and lateral material property gradient indexes, the peak of the oscillation amplitude and its associated excitation frequency increase.

Analytical mathematical solution for vibrational response of postbuckled laminated FG-GPLRC nonlocal strain gradient micro-/nanobeams

An inhomogeneous beam model incorporating size effects is formulated based on the nonlocal strain gradient theory of elasticity within the framework of a third-order shear deformation beam theory to analyze the size dependency in vibration behavior of postbuckled laminated functionally graded (FG) micro-/nanobeams made from graphene platelet-reinforced composite (GPLRC). The graphene platelets are randomly dispersed in each individual layer in such a way that the weight fraction of the nanofiller varies layerwise on the basis of different patterns of FG dispersion. Based upon the Halpin–Tsai micromechanical scheme, the effective material properties of laminated FG-GPLRC micro-/nanobeams are extracted. With the aid of the Hamilton’s principle, the nonlocal strain gradient equations of motion are constructed. After that, an improved perturbation technique is put to use to capture the small-scale effects on the fundamental frequencies of laminated FG-GPLRC nonlocal strain gradient micro-/nanobeams. The size-dependent natural frequencies of laminated third-order shear deformable FG-GPLRC micro-/nanobeams are obtained as function of applied axial compressive load within both the prebuckling and postbuckling domains. It is seen that in the prebuckling regime, the softening-stiffness effect of the nonlocality causes to reduce the natural frequencies of laminated FG-GPLRC micro-/nanobeam, but the hardening-stiffness influence of strain gradient size effect leads to increase the natural frequencies. However, by moving to the postbuckling regime, this pattern becomes vice versa, as the nonlocal size dependency leads to increase the natural frequencies of micro-/nanobeam while the strain gradient size effect causes to reduce them.

Processing relative clauses across comprehension and production: similarities and differences

ABSTRACTWe compare the processing of relative clauses in comprehension (self-paced reading) and production (planned production). We manipulated the locality of two syntactic dependencies: filler-gap (subject vs object gap) and subject–verb (centre-embedded vs right-branched). The non-local filler-gap dependency resulted in a longer embedded predicate duration, across domains, consistent with memory-based accounts. For the non-local subject–verb dependency, we observe longer reading times at the main verb, but in production a greater likelihood and duration of a pause preceding the main verb. We argue that this result stems from the cost of computing the restriction, which manifests as a prosodic break. In the context of the subject–verb dependency manipulation, we also revisit the source of interpretation break-down in multiple centre-embedding. Generally, our findings imply that memory-based accounts are adequate for filler-gap, but not subject–verb, dependencies and production studies can aid in understanding complexity effects.

Buckling and vibration analysis of nonlocal axially functionally graded nanobeams based on Hencky-bar chain model

• We introduce Hencky bar net model for inhomogenous nanobeams for the first time. • We show the formal analogy between Hencky bar net model and Finite Difference Model. • We demonstrate the convergence to continuum plate results by increasing the grid size for any proposed example. • The model deals with both stepped and tapered microbeams. Buckling and free vibration analyses of nonlocal axially functionally graded Euler nanobeams is the main objective of this paper. Due to its simplicity, the Eringen's differential constitutive model is adopted for describing the nonlocal size dependency of nanostructure beam. The nonlocal equilibrium equation is derived using the principle of the minimum potential energy principle, and discretized by using the link-spring model known in literature as Hencky bar-chain model. The general applicability of the proposed approach allows analyses of functional graded microbeams without any restriction on variability, boundary and loading conditions. A comparison with results available in the literature shows the reliability of the method.

Second Language Processing of Korean Floating Numeral Quantifiers.

This study investigated whether Chinese-Korean bilinguals can use structure-based information to interpret Korean sentences containing floating numeral quantifiers during online processing. A numeral quantifier in Korean can be stranded from its modified noun through scrambling as long as the quantifier forms a constituent with the noun. For Chinese-Korean bilinguals, acquiring this structural knowledge gives rise to a learnability problem because this ability cannot be derived from the L1, not easily induced from Korean input and is not obtained through classroom instruction. In acceptability judgment, highly proficient Chinese-Korean bilinguals demonstrated target-like knowledge of this structural constraint. Results from a self-paced reading task showed that the bilinguals showed sensitivity to the violation of the mismatch between an NQ and its modified NP, both in local and non-local dependency conditions. Our findings suggest that structure-based processing is possible for highly proficient bilinguals even when the target structure is not instantiated in bilinguals' L1.

Small scale effects on buckling and postbuckling behaviors of axially loaded FGM nanoshells based on nonlocal strain gradient elasticity theory

By means of a comprehensive theory of elasticity, namely, a nonlocal strain gradient continuum theory, size-dependent nonlinear axial instability characteristics of cylindrical nanoshells made of functionally graded material (FGM) are examined. To take small scale effects into consideration in a more accurate way, a nonlocal stress field parameter and an internal length scale parameter are incorporated simultaneously into an exponential shear deformation shell theory. The variation of material properties associated with FGM nanoshells is supposed along the shell thickness, and it is modeled based on the Mori-Tanaka homogenization scheme. With a boundary layer theory of shell buckling and a perturbation-based solving process, the nonlocal strain gradient load-deflection and load-shortening stability paths are derived explicitly. It is observed that the strain gradient size effect causes to the increases of both the critical axial buckling load and the width of snap-through phenomenon related to the postbuckling regime, while the nonlocal size dependency leads to the decreases of them. Moreover, the influence of the nonlocal type of small scale effect on the axial instability characteristics of FGM nanoshells is more than that of the strain gradient one.

Flux Noise in a Superconducting Transmission Line

We study a superconducting transmission line (TL) formed by distributed LC oscillators and excited by external magnetic fluxes which are aroused from random magnetization (A) placed in substrate or (B) distributed at interfaces of a two-wire TL. Low-frequency dynamics of a random magnetic field is described based on the diffusion Langevin equation with a short-range source caused by (a) random amplitude or (b) gradient of magnetization. For a TL modeled as a two-port network with open and shorted ends, the effective magnetic flux at the open end has non-local dependency on noise distribution along the TL. The flux-flux correlation function is evaluated and analyzed for the regimes (Aa), (Ab). (Ba), and (Bb). Essential frequency dispersion takes place around the inverse diffusion time of random flux along the TL. Typically, noise effect increases with size faster than the area of TL. The flux-flux correlator can be verified both via the population relaxation rate of the qubit, which is formed by the Josephson junction shunted by the TL with flux noises, and via random voltage at the open end of the TL.

Nonlocal size dependency in nonlinear instability of axially loaded exponential shear deformable FG-CNT reinforced nanoshells under heat conduction

The present study deals with size-dependent nonlinear instability characteristics of functionally graded carbon nanotube (FG-CNT) reinforced composite shells at nanoscale subjected to axial compression combined with through-thickness heat conduction. To take size dependency into account, Eringen’s nonlocal continuum elasticity is incorporated to a novel shear deformation shell theory including a refined exponential distribution for transverse shear strain. In addition to the uniform distribution (UD) of CNT reinforcements, three FG patterns are also considered, namely FG-A, FG-V and FG-X. Also, on the basis of polynomial series, the temperature variation due to the through-thickness heat conduction is estimated. Via a perturbation-based boundary layer-type solving procedure, explicit expressions for nonlocal equilibrium curves are proposed relevant to the prebuckling and postbuckling regimes of FG-CNT exponential shear deformable nanoshells with temperature-dependent and temperature-independent material properties. It is observed that by taking the nonlocality size effect into consideration, the influence of the through-thickness heat conduction on the nonlinear axial instability response of FG-CNT reinforced nanoshells becomes more significant.

The matching effect in resumption: A local analysis based on Case attraction and top-down derivation

In this paper we analyze a hitherto unstudied matching effect in resumptive relatives. In some languages where gaps and resumptives are in complementary distribution, the choice between the two strategies depends on the Case of the head noun: in Swiss German, the focus of our study, dative relativization requires resumptives; however, the resumptive is omitted if the head noun bears dative as well. This non-local dependency poses a serious challenge to local derivational bottom-up theories of syntax. We argue that a local solution is possible if the distribution of gaps and resumptives is reinterpreted in terms of Case attraction and the derivation unfolds top-down. Consequently, the relevant piece of information, the Case of the head noun, is available on the operator so that the choice between gaps and resumptives can be made without recourse to non-local devices. Gap relatives obtain in configurations where the Case attraction derivation converges while resumptives occur as a repair in derivations where Case attraction leaves a Case-probe unchecked. The matching effect falls out naturally as a subcase of Case attraction.

Constructing covert dependencies—The case of Mandarin wh-in-situ dependency

Abstract Wh-in-situ constructions in Mandarin Chinese, as opposed to their English counterparts that front wh-phrases to the beginning of the sentence, have the same word order as regular non-wh declaratives. We argue that despite their surface word order, processing wh-in-situ constructions involves constructing a covert non-local syntactic dependency between the in-situ wh-phrase and a higher scope position at a clause boundary, leading to behavioral patterns similar to those associated with the processing of overt dependencies. In two comprehension experiments, we showed that the process of linking an in-situ wh-phrase and its scope position induces a similarity-based memory interference effect if another clause boundary position intervenes. In addition, a set of sentence completion studies also showed that the production of wh-in-situ constructions is heavily modulated by the increased working memory burden that results from planning and maintaining a non-local dependency.

Direct Evidence of Memory Retrieval as a Source of Difficulty in Non‐Local Dependencies in Language

Linguistic dependencies between non-adjacent words have been shown to cause comprehension difficulty, compared with local dependencies. According to one class of sentence comprehension accounts, non-local dependencies are difficult because they require the retrieval of the first dependent from memory when the second dependent is encountered. According to these memory-based accounts, making the first dependent accessible at the time when the second dependent is encountered should help alleviate the difficulty associated with the processing of non-local dependencies. In a dual-task paradigm, participants read sentences that did or did not contain a non-local dependency (i.e., object- and subject-extracted cleft constructions) while simultaneously remembering a word. The memory task was aimed at making the word held in memory accessible throughout the sentence. In an object-extracted cleft (e.g., It was Ellen whom John consulted…), the object (Ellen) must be retrieved from memory when consulted is encountered. In the critical manipulation, the memory word was identical to the verb's object (ELLEN). In these conditions, the extraction effect was reduced in the comprehension accuracy data and eliminated in the reading time data. These results add to the body of evidence supporting memory-based accounts of syntactic complexity.