Search

Deep neural networks (DNNs) are prone to the notorious catastrophic forgetting problem when learning new tasks incrementally. Class-incremental learning (CIL) is a promising solution to tackle the challenge and learn new classes while not forgetting old ones. Existing CIL approaches adopted stored representative exemplars or complex generative models to achieve good performance. However, storing data from previous tasks causes memory or privacy issues, and the training of generative models is unstable and inefficient. This paper proposes a method based on multi-granularity knowledge distillation and prototype consistency regularization (MDPCR) that performs well even when the previous training data is unavailable. First, we propose to design knowledge distillation losses in the deep feature space to constrain the incremental model trained on the new data. Thereby, multi-granularity is captured from three aspects: by distilling multi-scale self-attentive features, the feature similarity probability, and global features to maximize the retention of previous knowledge, effectively alleviating catastrophic forgetting. Conversely, we preserve the prototype of each old class and employ prototype consistency regularization (PCR) to ensure that the old prototypes and semantically enhanced prototypes produce consistent prediction, which excels in enhancing the robustness of old prototypes and reduces the classification bias. Extensive experiments on three CIL benchmark datasets confirm that MDPCR performs significantly better over exemplar-free methods and outperforms typical exemplar-based approaches. • Multi-granularity knowledge distillation is proposed to alleviate catastrophic forgetting for class-incremental classification. • Prototype consistency regularization boosts the performance of CIL significantly and effectively mitigates classification bias. • Our framework performs well even when the previous training data is unavailable. [ABSTRACT FROM AUTHOR]

The Chinese Visceral Adiposity Index (CVAI) is a reliable indicator of visceral adiposity dysfunction in the Chinese population. We aimed to evaluate the association between CVAI and clinical outcome in Chinese ischemic stroke patients who received endovascular thrombectomy (EVT). This study retrospectively included patients with large vessel occlusive stroke receiving EVT treatment in 2 China stroke centers. Baseline CVAI was calculated after admission. Patients with a modified Rankin scale score ≥ 3 at 3 months after ischemic stroke were defined as poor outcome. Binary multivariate logistic regression models were utilized to explore the association between CVAI and the risk of 90-day unfavorable outcome. A total of 453 patients (mean age, 70.4 ± 12.1 years; 280 male) were included. During the 90-day follow-up, 236 (52.1 %) patients experienced poor outcome. After multivariable adjustment for potential confounders, increasing CVAI was associated with an increased risk of 90-day poor outcome (odds ratios, per-standard deviation increase: 1.521; 95 % confidence interval, 1.127−2.052; P = 0.006). Similar significant results were observed when the CVAI was analyzed as a categorical variable. Furthermore, the multiple-adjusted spline regression model showed an inverted J-shape association between CVAI and risk of unfavorable outcome (P = 0.048 for non-linearity). This study demonstrated that CVAI is positively correlated with 90-day poor outcome in Chinese ischemic stroke patients after EVT. [ABSTRACT FROM AUTHOR]

Abstract In a wireless power transfer (WPT) system, the transfer performance is related to the mutual inductance between coils. However, the mutual inductance decreases with the increase of transfer distance. In this work, the relationship between the output voltage and the mutual inductance for WPT systems with air core and with ferrite core are analyzed. In order to improve the mutual inductance, a novel configuration of receiving resonator with a strong magnetic coupling is proposed. The mutual inductance and magnetic field distribution for coils with a cylindrical core and with the novel configuration are compared. Experiments are carried out for validation. The results indicate that the proposed WPT system is superior to the system with the cylindrical ferrite core in increasing the output voltage and power transfer efficiency. [ABSTRACT FROM AUTHOR]

Abstract In this paper, a novel compact broadband rectifying antenna (rectenna) is presented for ambient radio frequency (RF) energy harvesting. Slot antennas with excellent bandwidth and matching characteristic are designed by etching fractal geometries with different iteration order on a radiation patch. A rectifier with a single stub matching network is investigated for impedance matching and the RF to DC conversion efficiency is improved at low input power. Simulation and the experiments have been carried out to evaluate the performance of antenna and rectifier respectively. A good match is observed between the simulated and measured results. The experimental results show that antenna features a broad bandwidth from 2.15 GHz to 2.9 GHz and the maximum conversion efficiency of rectifier is up to 64% at 0 dBm. The highest conversion efficiency of rectenna is 62% and the maximum output voltage is 1.5 V at the input power of 0 dBm. [ABSTRACT FROM AUTHOR]

• A fully connected neural network-based voltage compensation method is proposed for accurate reconstruction in brain electrical impedance tomography. • Under the condition of simulated dehydration of brain tissue and scalp, dehydration degree can be determined and voltage can be compensated. • The influence of the scalp dehydration is reduced and reconstruction quality is improved when the proposed compensation method is used. Acute cerebral ischemia is commonly accompanied by cerebral edema. With brain electrical impedance tomography (EIT), it is possible to monitor conductivity variation during dehydration treatment of the cerebral edema. However, the scalp is also dehydrated when dehydrating brain tissue which exerts a large impact on boundary measurement. Consequently, reconstruction quality of cerebral edema is seriously affected. To reduce the influence of scalp dehydration, a voltage data compensation method based on a fully connected neural network (FCNN) is proposed in this work. With this method, the voltage measured when brain tissue and scalp are simultaneously dehydrated can be compensated. To evaluate the performance of the proposed method, reconstruction of simulated cerebral ischemia in a three-layer head model is conducted. Images reconstructed with the compensated voltage data are compared with the results when there is no compensation. Besides, quantitative comparison is performed. The results indicate that it is almost impossible to identify the simulated cerebral ischemia without voltage compensation. In contrast, cerebral ischemia can be much more accurately reconstructed when the measured voltage is compensated with the proposed method. [ABSTRACT FROM AUTHOR]

Precisely controlling grain boundary (GB) thermodynamic stability and strength is important for processing nanograined alloys. The segregation-induced stabilization and strengthening GB mechanisms of 20 transition metal elements in Ni Σ11 [110](113) STGB with low interfacial excess energy were clarified from first-principles total-energy calculations in this work. The results indicate that the segregation energy and GB energy are inversely related to the radius difference between TM and matrix Ni atoms. The segregation of 11 elements were identified to enable the Ni Σ11 [110](113) STGB to be stable according to Schuh criterion for GB stability. The electronic mechanism of GB stabilization through TM segregation is mainly due to the less anti-bonding states of Ni -X pairs than those of Ni–Ni pairs in unsegregated GB, from the analysis of crystal orbital Hamiltonian populations. The GB embrittlement resulted from TM segregation is mainly from the mechanical contribution caused by atomic size mismatch between segregation atom and matrix atom. The segregation caused GB strengthening comes from the chemical contribution, which is due to the fact that the segregation of X leads to the formation of stronger Ni -X covalent-like bonds instead of Ni–Ni metallic bonds. This work could provide better understanding on designing high-stable Ni nanograined alloys. • The elements that stabilize grain boundaries are screened by Schuh criterion. • The grain boundary's cohesion was characterized by first-principles tensile tests. • Stabilizing and strengthening mechanisms were uncovered at electronic level. [ABSTRACT FROM AUTHOR]

Abstract: Wang proposed a gradient-based neural network (GNN) to solve online matrix-inverses. Global asymptotical convergence was shown for such a neural network when applied to inverting nonsingular matrices. As compared to the previously-presented asymptotical convergence, this paper investigates more desirable properties of the gradient-based neural network; e.g., global exponential convergence for nonsingular matrix inversion, and global stability even for the singular-matrix case. Illustrative simulation results further demonstrate the theoretical analysis of gradient-based neural network for online matrix inversion. [Copyright &y& Elsevier]

Abstract: Potential noise records have been collected for 2024-T3 aluminum alloy, which was exposed to simulated acid rain with different pH value for 15 wet–dry cycles. Meanwhile, Potentiodynamic polarization and SEM techniques were also used as assistant measurements. Three mathematic methods including average, standard deviation and wavelet transformation have been employed to analyze the records. The results showed that each single wet–dry cycle can be divided into three regions with respect to the change of the cathodic reaction rate, and with the increase of pH value the main cathodic reaction changes from the reduction of protons to that of oxygen molecules. The analysis of the EDP versus time evolution clearly indicates that the whole corrosion process can be divided into three segments for the case of pH 3.5 and only one for the cases of pH 4.5 and 6.0, which have been theoretically interpreted according to the corrosion theory and experimentally proved by SEM. The results also showed that the corrosion in the case of pH 3.5 was much more rigorous than that in the cases of pH 4.5 and 6.0. It may due to synergistic effects of that, the characteristic of hydrogen ions which is much more active than that of oxygen molecules, the high diffusion/migration rate of hydrogen ions in solution or through surface films and the lower stability of surface passive film at low pH value system. [Copyright &y& Elsevier]

Compared with electrical resistance tomography, capacitively coupled electrical resistance tomography (CCERT) is preferred since it avoids problems of electrode corrosion and electrode polarization. However, reconstruction of conductivity distribution is still a great challenge for CCERT. To improve reconstruction quality, this work proposes a novel image reconstruction method based on total fractional-order variation regularization. Simulation work is conducted and reconstruction of several typical models is studied. Robustness of the proposed method to noise is also conducted. Additionally, the performance of the proposed reconstruction method is quantitatively evaluated. We have also carried out phantom experiment to further verify the effectiveness of the proposed method. The results demonstrate that the quality of reconstruction has been largely improved when compared with the images reconstructed by Landweber, Newton-Raphson and Tikhonov methods. The inclusion is more accurately reconstructed and the background is much clearer even under the impact of noise. • A novel image reconstruction method based on total fractional-order variation regularization is proposed for CCERT. • Various simulation models and experimental cases have been accurately reconstructed. • Reconstruction based on the proposed method shows strong robustness to noise. [ABSTRACT FROM AUTHOR]

The image reconstruction of conductivity distribution in electrical impedance tomography (EIT) is a seriously ill-posed inverse problem. To cope with the problem, it is recognized that the regularization method is an effective approach. In this paper, an adaptive non-convex hybrid total variation (ANHTV) regularization method is proposed to reconstruct the conductivity distribution in EIT. The iterative reweighted least squares algorithm and the iterative alternating direction method of multipliers algorithm are developed to solve the ANHTV-based inverse model in the image reconstruction. Besides, all the parameters utilized in the inverse model are adaptively selected. To validate the advantage of the proposed method, extensive numerical simulation and experimental work have been carried out. Also, qualitative and quantitative comparisons with two convex TV-based regularization methods are conducted. The results show that the proposed method is more advantageous in terms of staircase effect suppression, edge information preservation and noise resisting in the image reconstruction. • A non-convex hybrid total variation regularization method is proposed for image reconstruction in EIT. • All the essential parameters in the proposed method are adaptively selected. • The staircase effect is suppressed and the edge information is preserved. [ABSTRACT FROM AUTHOR]

Electrical resistance tomography (ERT) is a promising technique with which the conductivity distribution in the detected region can be visualized. Mathematically, the reconstruction of conductivity distribution is a seriously ill-posed inverse problem which poses a great challenge for the ERT sensing technique. The regularization method has been found to be an effective approach in coping with the inverse problem. In this work, a novel reconstruction strategy which combines the non-convex regularization method with Landweber method is proposed for the image reconstruction in ERT. At each iteration, the non-convex regularization is used to constrain the conductivity calculated with the Landweber method. A simple and efficient generalized iterated shrinkage algorithm is developed to solve the proposed method. To validate the performance of the proposed method, a series of numerical simulation is conducted and comparative analysis with other methods is performed. From the results, it can be observed that images with high quality are obtained when reconstructing with the proposed method. The impact of noise on the reconstruction is also investigated which shows that the images reconstructed by the proposed method are the least sensitive to the noise. The performance of the proposed method in the image reconstruction is also verified by experimental data. The results demonstrate that the inclusion is accurately reconstructed and the background is clear when the proposed method is adopted for the image reconstruction. • A non-convex regularization method combined with Landweber method is proposed for electrical resistance tomography. • Inclusion are accurately reconstructed and the background is clear. • The reconstruction is robust to the noise. [ABSTRACT FROM AUTHOR]

MoSi 2 -type tetragonal iron dialuminide (space group I4/mmm) have been successfully synthesized and determined in a previous stud by Tobita et al. In this work, the electronic structure, effective mass , vibration, and electromagnetic properties of t -FeAl 2 were predicted using the density functional theory (DFT). First, the predicted crystal lattice parameters reproduced the previously reported experimental data. Second, the electronic structure shows that t -FeAl 2 is a narrow-gap semiconductor material with an indirect gap (0.059 eV), and the effective masse of the electron (m e *) and hole (m h *) is 0.35 m 0 and 0.2 m 0 , respectively. Third, the electromagnetic wave absorption characteristics such as the frequency-dependence of the relative complex permittivity, dielectric loss tangent tg δ , electrical conductivity, quality factor Q × f and reflection loss are calculated. Besides, using the linear response theory, the phonon dispersion curves and vibrational properties are calculated. The phonon dispersion curves do not have imaginary phonon frequencies, confirming the dynamic stability of t -FeAl 2. Finally, the vibrational properties are investigated, and the infrared (IR)-active and Raman-active modes are also assigned and briefly discussed. [ABSTRACT FROM AUTHOR]

In unsupervised domain adaptive object detection, learning target-specific features is pivotal in enhancing detector performance. However, previous methods mostly concentrated on aligning domain-invariant features across domains and neglected integrating the specific features. To tackle this issue, we introduce a novel feature learning method called Joint Feature Differentiation and Interaction (JFDI), which significantly boosts the adaptability of the object detector. We construct a dual-path architecture based on we proposed feature differentiate modules: One path, guided by the source domain data, utilizes multiple discriminators to confuse and align domain-invariant features. The other path, specifically tailored to the target domain, learns its distinctive characteristics based on pseudo-labeled target data. Subsequently, we implement an interactive enhanced mechanism between these paths to ensure stable learning of features and mitigate interference from pseudo-label noise during the iterative optimization. Additionally, we devise a hierarchical pseudo-label fusion module that consolidates more comprehensive and reliable results. In addition, we analyze the generalization error bound of JFDI, which provides a theoretical basis for the effectiveness of JFDI. Extensive empirical evaluations across diverse benchmark scenarios demonstrate that our method is advanced and efficient. [ABSTRACT FROM AUTHOR]

To wirelessly harvest RF energy at LTE band, a novel compact and efficient dual-port microstrip rectenna is proposed. An antenna designed with two ports and featured by arrow-shaped, circular and rectangular slots is presented. A broadband rectifier including a voltage doubler rectifying circuit and a novel matching network is developed to convert the RF power captured by the antenna into DC power. For well matching between the antenna and the rectifying circuit, the matching network which consists of double 70-deg fan-shaped stubs, no-uniform transmission line and double 55-deg fan-shaped stubs is designed. The prototypes of the proposed antenna and rectifier are respectively fabricated. The simulation and experimental results show that the antenna is able to harvest energy from different frequencies when fed by separate microstrip feeding line. The distribution of current density along with the radiation pattern in E-plane and H-plane is investigated. Besides, the conversion efficiency and output voltage of the rectifier are studied. To test the performance of the proposed rectenna, an indoor measurement is conducted. The results demonstrate that the RF-to-DC conversion efficiency of the proposed antenna is around 42% and the output voltage is around 0.9 V, indicating its potential application in the RF energy harvest. [ABSTRACT FROM AUTHOR]

[Display omitted] • A combination of AOP, RDF, MSD, and coordination number was used for the first time to reveal the methane hydrate decomposition. • Among glycine, alanine, valine, serine, and glutamic acid, glycine had the fastest decomposition rate. • 5 mol% glycine showed the best efficiency for the hydrate decomposition. • Temperature and pressure have considerable effects on the decomposition of methane hydrate. Controllable mining is a prerequisite for commercial hydrate mining and the addition of chemicals makes the mining process more efficient. However, traditional chemical additives are environmentally harmful and give rise to damage to the environment. Amino acids are green additives, and their effect on the decomposition of hydrates remains to be unequivocally understood. To reveal the influencing mechanism of amino acids on the decomposition of methane hydrates, molecular dynamics (MD) simulations were performed to systematically investigate the influences of the amino acid types, concentrations, temperature, and pressure on hydrates decomposition. The results demonstrated that glycine had shown excellent performance on accelerating the decomposition of hydrates, followed by valine, serine, alanine, and glutamic acid. 5 mol% glycine was most conducive to the decomposition of hydrates. Temperature was a favorable factor for the decomposition of methane hydrate, and the decomposition rate had a nonlinear relationship with the increase in temperature. Additionally, there were two factors affecting the decomposition of hydrates under varying pressure conditions. The first was the inhibition of pressure on the decomposition of hydrates, called inhibiting effect. The second was that pressure had an effect on the solubility of amino acids which could promote the decomposition of hydrates, called the facilitation effect. The synergistic effect of inhibiting effect and facilitation effect determined the final hydrate decomposition rate. The obtained results provided a qualitative and quantitative analysis of the decomposition process of the amino acid-hydrate system on the molecular scale, and also laid a basis for the selection and design of novel amino acid molecules promoting hydrates decomposition. [ABSTRACT FROM AUTHOR]

The resonant frequency of coils in a four-coil magnetic coupling resonance wireless power transfer (MCR-WPT) system has a direct impact on transfer coefficient. For efficient power transfer, most of the systems operate at high-Q resonant state, causing a narrow bandwidth and reducing the system tolerance to resonant frequency deviation. To deal with the problem, a four-coil MCR-WPT system characterized with broadband performance is investigated. Based on the equivalent circuit, the transfer coefficient and the input impedance are firstly calculated. The transfer characteristic at different resonant frequency under different transfer distances is discussed. By comparing the number and variance of the extreme values of transfer coefficient under different resonant frequencies, the resonant frequency for excellent broadband transmission is determined. To validate the broadband performance, frequency drift caused by capacitance variations is investigated at the resonant frequency of 19 MHz under the transfer distance of 45 mm. The results indicate that the MRC-WPT system generally maintains broadband transmission and the transfer coefficient at the excitation frequency of 19 MHz remains above 0.9 when the resonant frequency is deviated. Experiments have been performed to validate the excellent broadband performance. The results are generally consistent with the simulation analysis. [ABSTRACT FROM AUTHOR]