Search

Graph convolutional networks (GCNs) successfully generalize convolutional neural networks to handle the graphs with high-order arbitrary structures. However, most existing GCNs variants consider only the local geometry of row vectors of high-dimensional data via example graph Laplacian, while neglecting the manifold structure information of column vectors. To address this problem, we propose the example-feature graph convolutional networks (EFGCNs) for semi-supervised classification. Particularly, we introduce the definition of the spectral example-feature graph (EFG) convolution that simultaneously utilizes the example graph Laplacian and feature graph Laplacian to better preserve the local geometry distributions of data. After optimizing the spectral EFG convolution with the first-order approximation, a single-layer EFGCNs is obtained. It is then further extended to build a multi-layer EFGCNs. Extensive experiments on remote sensing and citation networks datasets demonstrate the proposed EFGCNs show superior performance in semi-supervised classification compared with state-of-the-art methods. [ABSTRACT FROM AUTHOR]

Transformation-induced plasticity (TRIP) high-entropy alloys (HEAs) have drawn great attention as they present excellent mechanical properties, and their phase stability is critical for the underlying deformation mechanisms and the application temperature range. In this study, the kinetic phase transformation behavior of a dual-phase TRIP-HEA Fe50Mn30Co10Cr10 (at. %) was probed by in situ neutron diffraction during thermal cycling between 293 and 425 K. Continuous austenitic and martensitic transformation were visualized through the evolution of hexagonal close-packed phase fraction during thermal cycling. Specifically, thanks to the magnetic sensitivity of neutron diffraction, it was found that the martensitic transformation under cooling became suppressed when the antiferromagnetic ordering started at ∼326 K. This temperature was further confirmed as the Néel temperature by magnetization measurements. Thermodynamic calculations revealed that the suppression effect on martensitic transformation is attributed to the consumption of the chemical driving force by the magnetic ordering. The magnetic ordering at such relatively high temperature is associated with the high Mn content. These findings shed light on a potential strategy to achieve better mechanical properties of Mn-containing alloys by manipulating the magnetic property through tuning the Mn content. [ABSTRACT FROM AUTHOR]

Deep learning has shown superiority to extract more representative features from multimedia data in recent years. Recently, the most typical graph convolutional networks (GCN) has achieved excellent performance in the semi-supervised framework-based data representation learning tasks. GCN successfully generalizes traditional convolutional neural networks to encode arbitrary graphs by exploiting the graph Laplacian-based sample structure information. However, GCN only fuses the static structure information. It is difficult to guarantee that its structure information is optimal during the training process and applicable for all practical applications. To tackle the above problem, in this paper, we propose a manifold regularized dynamic graph convolutional network (MRDGCN). The proposed MRDGCN automatically updates the structure information by manifold regularization until model fitting. In particular, we build an optimization convolution layer formulation to acquire the optimal structure information. Thus, MRDGCN can automatically learn high-level sample features to improve the performance of data representation learning. To demonstrate the effectiveness of our proposed model, we apply MRDGCN on the semi-supervised classification tasks. The extensive experiment results on human activity datasets and citation network datasets validate the performance of MRDGCN compared with GCN and other semi-supervised learning methods. [ABSTRACT FROM AUTHOR]

Currently, the representation learning of a graph has been proved to be a significant technique to extract graph structured data features. In recent years, many graph representation learning (GRL) algorithms, such as Laplacian Eigenmaps (LE), Node2vec and graph convolutional networks (GCN), have been reported and have achieved great success on node classification tasks. The most representative GCN fuses the feature information and structure information of data, which aims to generalize convolutional neural networks (CNN) to learn data features with arbitrary structure. However, how to exactly express the structure information of data is still an enormous challenge. In this paper, we utilize hypergraph p -Laplacian to preserve the local geometry of samples and then propose an effective variant of GCN, i.e. hypergraph p -Laplacian graph convolutional networks (HpLapGCN). Since hypergraph p -Laplacian is a generalization of the graph Laplacian, HpLapGCN model shows great potential to learn more representative data features. In particular, we simplify and deduce a one-order approximation of spectral hypergraph p -Laplacian convolutions. Thus, we can get a more efficient layer-wise aggregate rule. Extensive experiment results on the Citeseer and Cora datasets prove that our proposed model achieves better performance compare with GCN and p -Laplacian GCN (pLapGCN). [ABSTRACT FROM AUTHOR]

Few-shot learning aims to learn a classifier with more generalization capability from extremely limited labeled samples has drawn an increasing amount of attention in many areas. One typical work in this field is the transductive propagation network (TPN), which propagates labels by capturing the local geometry distribution information between data. However, TPN exploited only the manifold structure of feature extractor networks' high-layer semantic features (single-scale) distribution, while neglecting their local geometry distribution in the low-layer semantic features. This paper presents the adaptive multi-scale transductive information propagation (AMTIP) model to address this problem. Specifically, we introduce the multi-scale feature extractor networks to simultaneously learn samples' high-layer global semantic features and low-layer local semantic features. After acquiring the multi-scale features, we propose the adaptive multi-scale fusion networks to generate the adaptive semantic fusion features applying to different few-shot classification tasks. Finally, the adaptive semantic fusion features are applied to the label propagation model for few-shot image classification. Compared with the single-scale semantic features of TPN, AMTIP can better preserve the local geometry between the support set and query set data via the adaptive multi-scale semantic fusion features. Extensive experiments demonstrate the superiority of the proposed AMTIP compared to the state-of-the-art few-shot classification models. [ABSTRACT FROM AUTHOR]

Abstract: The evolutions of uniaxial tension properties and the ratcheting fatigue behavior of AZ31 magnesium alloy sheet pre-corroded in the phosphate buffered solution (PBS) stimulated physiological environment with increasing immersion time were investigated in this study. The corrosion behavior of AZ31 sheet was also studied. It was found that the corrosion rate decreased exponentially with immersion time until stabilized at about 0.05g/m2 h after 28 days immersion. The corrosion behavior was characterized by pitting corrosion and progressive pitting corrosion in the direction of pits depth was gradually inhibited. The Young's modulus and the elongation of AZ31 sheet decreased exponentially with increasing immersion time and became low levels with reductions of around 10% and 20% respectively at 28 days immersion. The ultimate strength, however, decreased no more than 5%. With increasing immersion time, both the ratcheting strain and ratcheting strain rate in the stable stage of ratcheting strain evolution tended to increase. 28 days degradation decreased the fatigue lives by a factor of about 2.5–5. The pre-corroded damage defined as life reduction after degradation on the basis of the Miner's linear damage rule (LDR) was found to increase exponentially with immersion time regardless of the loading conditions. A ratcheting fatigue life prediction model based on the Miner LDR, the Smith–Watson–Topper (SWT) parameter and the Basquin's equation was proposed and yielded good prediction for the pre-corroded AZ31 sheet. [Copyright &y& Elsevier]

• Combined pre-compression significantly enhances the mechanical properties of magnesium alloys. • Combined pre-compression significantly reduces the material's anisotropy. • Combined pre-compression improves fatigue performance and weakens rapid hardening during cyclic loading. The research investigated the microstructure, monotonic mechanical properties, and fatigue behaviors of ZK60 magnesium alloy treated with single-pass pre-compression and combined pre-compression. The study found that combined pre-compression significantly improved mechanical strength in both the transverse and rolling directions compared to the single-pass method. Due to the randomized texture and intricate interwoven grain structures, the combined pre-compression method significantly mitigated the anisotropy observed in samples treated with single-pass pre-compression. Both pre-compression techniques enhanced the fatigue life of the ZK60 magnesium alloy. Samples subjected to combined pre-compression exhibited fatigue performance comparable to those treated with single-pass pre-compression. Additionally, there was a clear negative correlation between the fatigue life of the three samples and the magnitude of the mean stress. The abundance of pre-existing dislocations in the combined pre-compression processed samples helped in reducing the trend towards cyclic hardening. In contrast to single-pass pre-compression, the combined pre-compression method effectively reduced twinning activities and promoted the participation of basal slip during cyclic loading, leading to a more symmetrical hysteresis loop. [ABSTRACT FROM AUTHOR]

BACKGROUND: The error magnitude is closely related to patient-specific dosimetry and plays an important role in evaluating the delivery of the radiotherapy plan in QA. No previous study has investigated the feasibility of deep learning to predict error magnitude. OBJECTIVE: The purpose of this study was to predict the error magnitude of different delivery error types in radiotherapy based on ResNet. METHODS: A total of 34 chest cancer plans (172 fields) of intensity-modulated radiation therapy (IMRT) from Eclipse were selected, of which 30 plans (151 fields) were used for model training and validation, and 4 plans including 21 fields were used for external testing. The collimator misalignment (COLL), monitor unit variation (MU), random multi-leaf collimator shift (MLCR), and systematic MLC shift (MLCS) were introduced. These dose distributions of portal dose predictions for the original plans were defined as the reference dose distribution (RDD), while those for the error-introduced plans were defined as the error-introduced dose distribution (EDD). Different inputs were used in the ResNet for predicting the error magnitude. RESULTS: In the test set, the accuracy of error type prediction based on the dose difference, gamma distribution, and RDD + EDD was 98.36%, 98.91%, and 100%, respectively; the root mean squared error (RMSE) was 1.45–1.54, 0.58–0.90, 0.32–0.36, and 0.15–0.24; the mean absolute error (MAE) was 1.06–1.18, 0.32–0.78, 0.25–0.27, and 0.11–0.18, respectively, for COLL, MU, MLCR and MLCS. CONCLUSIONS: In this study, error magnitude prediction models with dose difference, gamma distribution, and RDD + EDD are established based on ResNet. The accurate prediction of the error magnitude under different error types can provide reference for error analysis in patient-specific QA. [ABSTRACT FROM AUTHOR]

Skeleton‐based action recognition is an inspired yet challenging task in computer vision. Recently, the latest graph convolutional network (GCN), which generalises well‐established convolutional neural networks to non‐Euclidean structures, is proven to be highly successful for action recognition from body skeleton data. However, the GCN architecture has not been fully studied. In this work, a Fisher vector (FV) encoding based GCN architecture (FV‐GCN) is proposed, which exceeds the limitations of existing GCN‐based methods by combining the GCN model with FV encoding. A channel‐wise spatial–temporal aggregation function to preserve spatial–temporal information in the whole action clip and integrate it into the FV‐GCN architecture is also presented. Since FV is different from the GCN structure, this hybrid architecture that incorporates the advantages of both algorithms can discover complementary information of feature representation effectively. On two challenging human action datasets, kinetics, and NTU‐RGBD, improved performance is demonstrated over the baseline method, and the FV‐GCN is better or comparable to some state‐of‐the‐art methods. [ABSTRACT FROM AUTHOR]

Miniaturization of modern devices drives the exploration into how materials behave in small dimensions. However, the mechanical behaviors of micro-scale specimens under varied loading conditions are still challenging to characterize and predict. In this work, specimen size-dependent weakening effect was experimentally revealed in 316L stainless steel thin wires under tension with wire diameter decreasing from 200 μm down to 61 μm. The underlying mechanisms were primarily associated with surface grain softening and strain localization, which were largely dependent on the number of grains across wire diameter. Through an analog composite model with surface and interior grains that follows the Hall-Petch relation, the extent of surface grain softening was found to coincide with the flow stress reduction due to lack of boundary strengthening. Further using elasto-plastic self-consistent polycrystal modeling, the stress-strain behaviors of interior and surface grains were well described in analogy to those of bulk materials with similar-size grains and extremely large grains, respectively. Subsequently, the analog composite model well simulated the specimen size-dependent weakening effect under tension. This work provided more insights into the size effect exploration and proposed a practical model for describing size-dependent mechanical behaviors of small-scale materials to be applied to miniaturized devices. [ABSTRACT FROM AUTHOR]

Highlights • Oligo-crystalline 316L steel thin wires showed a "thinner is stronger" size effect. • Reduced wire diameter led to lower ratcheting strain under tension-torsion cycling. • Multiaxial ratcheting behavior of thin wires appeared loading path independent. Abstract The uniaxial tensile and multiaxial ratcheting behaviors of oligo-crystalline 316L stainless steel thin wires of two diameters (190 and 90 μm) with less than 3.5 grains across diameter were experimentally investigated. The size effect of "thinner is stronger" was observed on the yield strength, though the ultimate tensile strength and elongation were lower in the thinner wire. Under six differently combined axial stress and shear strain cycling conditions, the ratcheting strains were all found to be relatively suppressed in the thinner wire, accompanied by enhanced shear stress flow. Moreover, unlike in bulk materials, the multiaxial ratcheting strain in both thin wires tended to be loading path independent in the cases with the same shear strain amplitude, axial mean stress and stress amplitude. The mechanisms of these size effects were analyzed to be related to the surface dislocation source limitation and strain gradient hardening. [ABSTRACT FROM AUTHOR]

This work reports the low cycle fatigue behavior of a wrought 316LN stainless steel under different control modes at room temperature. Under symmetrical strain and stress cycling, the steel exhibits consistent loading-amplitude-dependent cyclic hardening/softening and fatigue life characteristics. Under asymmetrical stress cycling, the steel is significantly hardened due to mean stress, and the fatigue life at the same strain amplitude is significantly reduced due to ratcheting strain. With the increase of mean stress, though the ratcheting strain level is increased, the fatigue life is prolonged. The effect of mean-stress hardening and ratcheting strain on fatigue life is discussed in terms of strain amplitude and micro-crack initiation and propagation. The Smith-Walker-Topper (SWT) model and a newly proposed fatigue life model based on the Coffin-Manson equation were used to predict the fatigue life under mean stress, and the proposed model yields more robust predictions. [ABSTRACT FROM AUTHOR]

Multiaxial fatigue test were conducted on PEEK under four loading paths and several criteria were used for life prediction. Predictions based on Fatemi–Socie criterion fall out of the factor-of-two line on the conservative side for proportional loading and non-proportional loading. The normal strain-based Smith–Watson–Topper criterion yield good prediction for proportional loading but non-conservative prediction for non-proportional loading. Taking the contribution from shear component into consideration, the Chen–Xu–Huang criterion achieve satisfactory results for both proportional loading and non-proportional loading, but with non-conservative prediction for pure torsion loading. To consider general cracking behavior, the modified SWT criterion is adopted and good agreement between experimental data and prediction values is achieved. [ABSTRACT FROM AUTHOR]

Precise monitoring of crack growth and a comprehensive understanding of crack propagation rate are crucial for the evaluation of the structural integrity of magnesium alloys with defects. Different crack deflections caused by the coupling of specimen orientations and crack sizes significantly increase the challenge to magnesium alloy evaluation. In this paper, the systematic finite element analysis was conducted for compact tension (CT) specimens with deflection cracks. A wide range of geometric variations including initial crack length, crack deflection angle and crack propagation length were analyzed. Complete solutions of the stress intensity factor (K I and K II), T -stress and the compliance at the load line were determined. The results showed that the geometric sizes have a significant coupling effect on the fracture mechanics parameters, and the non-deflected cracked formula recommended by the ASTM standard is not applicable to the deflection crack specimens. Then, a prediction method of kinked fatigue crack length of CT specimen based on the compliance multiplier method was proposed. To further verify the importance of the fracture parameters obtained, CT fatigue crack growth (FCG) tests of ZK60 magnesium (Mg) alloy under different orientations and initial crack sizes were carried out. The FCG rate curves of ZK60 Mg alloy under different crack growth driving forces were compared and illustrated. The developed compliance multiplication method can accurately predict the real-time crack propagation length. • The finite element analysis was conducted for CT specimens with deflection cracks. • The K I , K II , T -stress and compliance were determined. • The coupling effects of crack length and kink angle on fracture parameters were analyzed. • Kinked crack length prediction based on compliance multiplier method was proposed. • FCG tests were conducted to apply and validate the obtained fracture parameters. [ABSTRACT FROM AUTHOR]