Search

Deep learning-based methods have achieved excellent performance in image-deraining tasks. Unfortunately, most existing deraining methods incorrectly assume a uniform rain streak distribution and a fixed fine-grained level. And this uncertainty of rain streaks will result in the model not being competent at repairing all fine-grained rain streaks. In addition, some existing convolution-based methods extend the receptive field mainly by stacking convolution kernels, which frequently results in inaccurate feature extraction. In this work, we propose momentum-contrast and large-kernel for multi-fine-grained deraining network (MOONLIT). To address the problem that the model is not competent at all fine-grained levels, we use the unsupervised dictionary contrastive learning method to treat different fine-grained rainy images as different degradation tasks. Then, to address the problem of inaccurate feature extraction, we carefully constructed a restoration network based on large-kernel convolution with a larger and more accurate receptive field. In addition, we designed a data enhancement method to weaken features other than rain streaks in order to be better classified for different degradation tasks. Extensive experiments on synthetic and real-world deraining datasets show that the proposed method MOONLIT achieves the state-of-the-art performance on some datasets. Code is available at https://github.com/awhitewhale/moonlit. [ABSTRACT FROM AUTHOR]

Snow is a harsh natural phenomenon that greatly affects the performance of advanced computer vision tasks. Recently, most image desnowing methods rely on complex model structures, leading to increased carbon emissions and impossibilities to deploy on lightweight computing devices. Additionally, most deep learning-based methods do not effectively utilize the position information of snow particles, which will limit the performance of the model. To address these issues, considering that snow particles have block-shaped shape distribution and color uniformity similar to a mask, we propose a snowed autoencoder (SAE) desnowing method. Specifically, the SAE is composed of four parts: snowed masking process, SAE encoder, SAE decoder, and prediction stage. First, a snowy image is passed through the snowed masking process to generate a mask that exactly covers the snow particles and output the coordinates of the four vertices of the mask and the index of each segmented image patch. The input image and each mask coordinate and index are passed through the SAE encoder based on the snow particle attention module to generate an image representation for identifying typical features of snow particles. Then, the generated typical features of snow particles are fed into the SAE decoder to reconstruct image patches without snow. Finally, all the patches will be passed through the prediction stage to reconstruct the original size of the snow-free image. A large number of experiments show that the proposed SAE desnowing method achieves the state-of-the-art desnowing performance on three synthetic and one real-world desnowing datasets. The images after desnowing by SAE have better color details and are more consistent with human visual habits. In addition, SAE has faster desnowing speed and fewer parameters. The code is available at https://github.com/awhitewhale/SAE. [Display omitted] • An efficient desnowing method named Snowed Autoencoder (SAE) is proposed. • A multi-encoder for SAE using snow information with an attention module is proposed. • A multi-decoder for SAE with fewer sub-decoders than sub-encoders is proposed. • SAE outperformed state-of-the-art on four desnowing datasets. [ABSTRACT FROM AUTHOR]

Visual scenes containing text in the TextVQA task require a simultaneous understanding of images, questions, and text in images to reason answers. However, most existing cross-modal tasks merely involve two modalities. There are thus few methods for modeling interactions across three modalities. To bridge this gap, we propose in this work cross- and intra-modal interaction modules for multiple (more than two) modalities, where scaled dot-product attention method is applied to model inter- and intra-modal relationship. In addition, we introduce guidance information to assist the attention method to learn a more accurate relationship distribution. We construct a Multi-level Complete Interaction (MLCI) model for the TextVQA task via stacking multiple blocks composed of our proposed interaction modules. We design a multi-level feature joint prediction approach to exploit output representations from each block in a complementary way to predict answers. The experimental results on the TextVQA dataset show that our model obtains a 5.42% improvement in accuracy more than the baseline. Extensive ablation studies are carried out for the comprehensive analysis of the proposed method. Our code is publicly available at https://github.com/zhangshengHust/mlci. [ABSTRACT FROM AUTHOR]

Osteoarthritis (OA) is a chronic inflammatory joint disease. Traditional chinese medicine provides a resource for drug screening for OA treatment. Ginseng and the associated bioactive compound, ginsenosides, may reduce inflammation, which is considered a risk factor for the development of OA. Specifically, ginsenosides may exhibit anti-inflammatory and anti-oxidative stress activities, and inhibit extracellular matrix degradation by suppressing the NF-κB and MAPK signaling pathways. Notably, specific ginsenosides, such as compound K and Rk1, may physically interact with IκB kinase and inhibit the associated phosphorylation. Thus, ginsenosides exhibit potential as therapeutic candidates in the management of OA. However, the low water solubility limits the clinical applications of ginsenosides. Numerous effective strategies have been explored to improve bioavailability; however, further investigations are still required. [ABSTRACT FROM AUTHOR]

A light delivery system based on localized surface plasmon (LSP) is the key component of heat-assisted magnetic recording (HAMR). We demonstrated a novel resonator–antenna system as the light delivery system for HAMR. We numerically analyzed the interaction mechanism between the resonator waveguide and the antenna. The localized enhanced resonated field of the resonator can improve the electric-field enhancement of the antenna. Based on the interaction mechanism, we proposed the ring resonator to achieve higher electric-field enhancement of light spots as the resonated field can be several times higher than input light under the resonance condition. The input laser energy can be recycled by limiting the energy in the ring waveguide without using higher power laser sources. Numerical results show that the nanofocusing performance in the ring resonator system would be much better than normal light systems. The maximum electric-field enhancement of light spots can be 124, and the absorption efficiency can be about 10% in the ring-lollipop system. The proposed resonator–antenna system can effectively increase the nanofocusing performance of the light delivery system for HAMR. [ABSTRACT FROM AUTHOR]

Two-dimensional magnetic recording (TDMR) by shingled magnetic recording (SMR) is one of the promising technologies to achieve a magnetic recording areal density (AD) of 10 Tb/in2. The readback signal of TDMR suffers severe 2-D inter-symbol interference (2-D ISI). At ultra-high AD, reader array is proposed to process the signals from the 2-D ISI channel because it can acquire multiple interfered signal streams simultaneously from the adjacent tracks. In this article, we propose a novel architecture for signal detection, which uses a four-reader array to recover the recording data. The auxiliary reader of the four-reader array can offer additional soft information of bits on the detected track, and this information can be used to improve the bit error rate (BER) performance. We consider three approaches to utilize the auxiliary reader and investigate the position of the auxiliary reader. Simulation results show that the proposed approach outperforms both the conventional single-reader detection and three-reader array detection. The auxiliary reader is set with an offset of 10% track pitch and can obtain a significant improvement of BER performance when the signal-to-noise rates (SNRs) of all the four readers are the same. [ABSTRACT FROM AUTHOR]

Methods based on convolutional neural networks have achieved excellent performance in the image dehazing task. Unfortunately, most of the dehazing methods that exist suffer from loss of detail in the convolution and activation operations and failure to consider the effects of superimposing different intensities of haze, such as under-exposed and over-exposed images. To address these issues, we propose a dynamic dehazing convolution (DDC) based on attentional weight calculation and dynamic weight fusion and a dynamic dehazing activation (DDA) based on the input global context encoding function to address the problem of detail loss. And we propose a multi-scaled feature-fused image dehazing network (MFID-Net) based on DDC and DDA to address the effects of haze superposition. We also design a loss function based on the physical model with dynamic weights. Extensive experimental results demonstrate that the proposed MFID-Net performs favorably against the state-of-the-art algorithms on the hazy dataset while improving further on hazy images with large differences in haze concentration, and producing satisfactory dehazing results. The code is available at https://github.com/awhitewhale/MFID-Net. • Multi-scaled feature-fused image dehazing network called MFID-Net is proposed. • MFID-Net uses dynamic dehazing convolution and activation to improve image dehazing. • The proposed dynamic dehazing convolution utilizes dynamic weight fusion. • The proposed dynamic dehazing activation uses input global context encoding function. • MFID-Net achieves the state-of-the-art performance on dehazing datasets. [ABSTRACT FROM AUTHOR]

A scale-similarity model of a two-point two-time Lagrangian velocity correlation (LVC) was originally developed for the relative dispersion of tracer particles in isotropic turbulent flows (HE, G. W., JIN, G. D., and ZHAO, X. Scale-similarity model for Lagrangian velocity correlations in isotropic and stationary turbulence. Physical Review E, 80, 066313 (2009)). The model can be expressed as a two-point Eulerian space correlation and the dispersion velocity V. The dispersion velocity denotes the rate at which one moving particle departs from another fixed particle. This paper numerically validates the robustness of the scale-similarity model at high Taylor micro-scale Reynolds numbers up to 373, which are much higher than the original values (Rλ = 66, 102). The effect of the Reynolds number on the dispersion velocity in the scale-similarity model is carefully investigated. The results show that the scale-similarity model is more accurate at higher Reynolds numbers because the two-point Lagrangian velocity correlations with different initial spatial separations collapse into a universal form compared with a combination of the initial separation and the temporal separation via the dispersion velocity. Moreover, the dispersion velocity V normalized by the Kolmogorov velocity Vη ≡ η/τη in which η and τη are the Kolmogorov space and time scales, respectively, scales with the Reynolds number Rλ as V/Vη∝Rλ1.39 obtained from the numerical data. [ABSTRACT FROM AUTHOR]

The shingled magnetic recording (SMR) with $\text{L}1_{0}$ -FePt-based exchange coupled composite (ECC) media is a promising way to achieve ultrahigh recording density. During the writing process, the insensitive angular dependence of switching field provided by ECC media is found to play an important role in mitigating transition noise and adjacent track erasure, furthermore the investigation indicates that the skew angle (SA) and corner angle (CA) of wide main pole inevitably affect the erase band noise in SMR. Hence, this paper studies an improved calculation method of erase bandwidth (EBW) to evaluate the erase band noise more accurately, and investigates the impact of SA and CA on write performance. Additionally, by investigating the bit error rate of adjacent track’s bits caused by the erasure, the results show that the high angular tolerance of switching field provided by $L1_{0}$ -FePt(6 nm)/[Co/Ni]5(4 nm) ECC media can effectively reduce EBW. Finally, the analyses demonstrate that an optimum combination of writer CA, tapered angle, and suitable range of SA is helpful to reduce erase band noise and improve the write performance. [ABSTRACT FROM AUTHOR]

The single-, two- and multi-particle dispersions in isotropic turbulent flows are investigated using the direct numerical simulation (DNS), filtered DNS (FDNS) and large-eddy simulation (LES) with a spectral eddy viscosity subgrid scale (SGS) model. The contributions of filtering operation and SGS model to the dispersions are separately studied by comparing the statistics obtained from the three methods. The missing SGS motions in LES can be observed to significantly hinder two-particle and four-particle dispersions if the initial separations are less than or comparable to the filter width. A theoretical analysis of the non-monotonic behavior at short time of the one-time, two-point Lagrangian velocity correlation functions with large initial separations based on the Taylor expansion and the Kolmogorov similarity theory is derived, and the Reynolds number effect on the performance of the spectral eddy viscosity SGS model is also investigated. The results show that the SGS model used performs better with increasing Reynolds numbers. It is concluded that the particle SGS model is needed to be developed to correctly capture the Lagrangian two- and multi-point dispersion statistics of fluid particles. [ABSTRACT FROM AUTHOR]

Traditionally, one single hard label determines the sentiment label of an entire utterance for speech sentiment analysis. It obviously ignores the inherent dynamic and ambiguity of speech sentiments. Moreover, there are few segment-level ground truth labels in the most existing sentiment corpora, due to the label ambiguity and annotation cost. In this work, to capture segment-level sentiment fluctuations across one utterance, we propose sentiment profiles (SPs) to express segment-level soft labels. Meanwhile, we introduce massive multi-modal wild video data to solve the data shortage problem, and facial expression knowledge is used to guide audio segments generate soft labels through the Cross-modal Sentiment Annotation Module. Then, we design a Speech Encoder Module to encode audio segments into SPs. We further exploit the sentiment profile purifier (SPP) to iteratively improve the accuracy of SPs. Numerous experiments show that our model achieves state-of-the-art performance on CH-SIMS and IEMOCAP datasets with unlabeled data respectively. [Display omitted] • To address the lack of high-quality large-scale labeled datasets for SSA, we introduce multi-modal wild video data to expand it. • To capture fine-grained sentiment cues at the segment level, we propose sentiment profiles (SPs) to express segment-level soft labels. • Facial expression knowledge is used to guide audio segments generate SPs through Cross-modal Sentiment Annotation Module. • We design a speech encoder module (SEM) to encode audio segments into SPs and a sentiment profile purifier (SPP) strategy to purify label credibility. [ABSTRACT FROM AUTHOR]

Particle-laden turbulent flows are ubiquitous in natural and engineering flows. The preferential concentration and pair-statistics of heavy particles at small and intermediate Stokes numbers are mainly related to the small-scale motions, which are missing in conventional large-eddy simulation (LES). Therefore, studies on the effects of subgrid scale (SGS) motions on particle-pair statistics are significant. This paper will study the effects of SGS motions on the collision-related statistics of a bidisperse system that involves two groups of particles at different Stokes numbers using a direct numerical simulation (DNS), filtered DNS (FDNS) and LES. Compared with a monodisperse system at a single Stokes number, the relative error of the collision rates in the bidisperse system is much smaller than that in the monodisperse one. For given filter widths and Stokes numbers, the relative errors between the FDNS and LES are much smaller than their relative errors to DNS, implying that the error caused by the filtering operation in LES plays a leading role in the overall error of the particle collision rate at Stokes numbers less than 3. A particle SGS model is, thus, necessary to consider the effect of SGS motions on bidisperse heavy particles. [ABSTRACT FROM AUTHOR]

The rotational motion and orientational distribution of ellipsoidal particles in turbulent flows are of significance in environmental and engineering applications. Whereas the translational motion of an ellipsoidal particle is controlled by the turbulent motions at large scales, its rotational motion is determined by the fluid velocity gradient tensor at small scales, which raises a challenge when predicting the rotational dispersion of ellipsoidal particles using large eddy simulation (LES) method due to the lack of subgrid scale (SGS) fluid motions. We report the effects of the SGS fluid motions on the orientational and rotational statistics, such as the alignment between the long axis of ellipsoidal particles and the vorticity, the mean rotational energy at various aspect ratios against those obtained with direct numerical simulation (DNS) and filtered DNS. The performances of a stochastic differential equation (SDE) model for the SGS velocity gradient seen by the particles and the approximate deconvolution method (ADM) for LES are investigated. It is found that the missing SGS fluid motions in LES flow fields have significant effects on the rotational statistics of ellipsoidal particles. Alignment between the particles and the vorticity is weakened; and the rotational energy of the particles is reduced in LES. The SGS-SDE model leads to a large error in predicting the alignment between the particles and the vorticity and over-predicts the rotational energy of rod-like particles. The ADM significantly improves the rotational energy prediction of particles in LES. [ABSTRACT FROM AUTHOR]

One of the feasible methods to achieve ultrahigh areal density of magnetic storage system is shingled recording with bit patterned media technology. The adjacent track erasure (ATE) may take place under the impact of the skew footprint, which is associated with the initial slider state, the skewed pole, and the writing corner angle. A sequence of shingled trapezoidal shaped writing heads is modeled to explore the skew effect versus the writing performance. The results show that larger corner angle and narrow skew range are recommended to enhance the writing field and avoid the ATE. The analysis also suggests that hot-spot disk data in the region where the skew angle ranges from −20° to 0° has the lowest possibility to be potentially disrupted. [ABSTRACT FROM AUTHOR]

The light delivery system with a light spot is the key component of heat-assisted magnetic recording (HAMR). We proposed a semi-circle resonator system as the light delivery system for HAMR. It consists of only a semi-circle resonator, a reflecting mirror, and a near-field transducer (NFT). The resonator can provide a stable enhanced optical field to excite the localized surface plasmon of NFT without complex focusing lenses or external controls. Numerical results show that the size of light spots can be 20–40 nm with a good field enhancement. Combined with bit-patterned media, the performance can be improved several times than the thin-film structure. The thermal profile of the light spot shows that the areal density of HAMR can further improve beyond the limit size of light spots, and the recording bit can be reduced to 12 × 12 nm2 corresponding to the areal density of about 4 T b / i n 2. • The semi-circle resonator system for HAMR is simple without complex focusing lenses or external controls. • The profile of light spots can be shrunk to 20-30 nm with a good field enhancement for the high areal density. • The potential areal density can further improve beyond the limit size of light spots in the bit-patterned HAMR. [ABSTRACT FROM AUTHOR]

VQA attracts lots of researchers in recent years. It could be potentially applied to the remote consultation of COVID-19. Attention mechanisms provide an effective way of utilizing visual and question information selectively in visual question and answering (VQA). The attention methods of existing VQA models generally focus on spatial dimension. In other words, the attention is modeled as spatial probabilities that re-weights the image region or word token features. However, feature-wise attention cannot be ignored, as image and question representations are organized in both spatial and feature-wise modes. Taking the question "What is the color of the woman's hair" for example, identifying the hair color attribute feature is as important as focusing on the hair region. In this paper, we propose a novel neural network module named "multimodal feature-wise attention module" (MulFA) to model the feature-wise attention. Extensive experiments show that MulFA is capable of filtering representations for feature refinement and leads to improved performance. By introducing MulFA modules, we construct an effective union feature-wise and spatial co-attention network (UFSCAN) model for VQA. Our evaluation on two large-scale VQA datasets, VQA 1.0 and VQA 2.0, shows that UFSCAN achieves performance competitive with state-of-the-art models. • Visual Question and Answering could be potentially applied to the remote consultation of COVID-19. • Feature-wise attention enables obtaining more discriminative representations. • Feature-wise attention is an effective complement for spatial attention. • Feature-wise attention could work in both image and question modalities. • Bilinear model performs better than element-wise multiplication in multimodal fusion. [ABSTRACT FROM AUTHOR]

Many deep learning (DL)-based image restoration methods for low-dose CT (LDCT) problems directly employ the end-to-end networks on low-dose training data without considering dose differences. However, the radiation dose difference has a great impact on the ultimate results, and lower doses increase the difficulty of restoration. Moreover, there is increasing demand to design and estimate acceptable scanning doses for patients in clinical practice, necessitating dose-aware networks embedded with adaptive dose estimation. In this paper, we consider these dose differences of input LDCT images and propose an adaptive dose-aware network. First, considering a large dose distribution range for simulation convenience, we coarsely define five dose levels in advance as lowest, lower, mild, higher and highest radiation dose levels. Instead of directly building the end-to-end mapping function between LDCT images and high-dose CT counterparts, the dose level is primarily estimated in the first stage. In the second stage, the adaptively learned low-dose level is used to guide the image restoration process as the pattern of prior information through the channel feature transform. We conduct experiments on a simulated dataset based on original high dose parts of American Association of Physicists in Medicine challenge datasets from the Mayo Clinic. Ablation studies validate the effectiveness of the dose-level estimation, and the experimental results show that our method is superior to several other DL-based methods. Specifically, our method provides obviously better performance in terms of the peak signal-to-noise ratio and visual quality reflected in subjective scores. Due to the dual-stage process, our method may suffer limitations under more parameters and coarse dose-level definitions, and thus, further improvements in clinical practical applications with different CT equipment vendors are planned in future work. [ABSTRACT FROM AUTHOR]

Background: Osteoporosis (OP), due to microarchitectural alterations, is associated with decreased bone mass, declined strength, and increased fracture risk. Increased osteoblast apoptosis contributes to the progression of OP. Natural compounds from herbs provide a rich resource for drug screening. Our previous investigation showed that geniposide (GEN), an effective compound from Eucommia ulmoides, could protect against the pathological development of OP induced by cholesterol accumulation. Methods: The rat OP models were duplicated. Dual-energy X-ray absorptiometry, hematoxylin and eosin staining, and immunohistochemistry were used to evaluate bone changes. TUNEL/DAPI staining assays were used for cell apoptosis detection. Protein expression was determined by western blotting assays. Results: A high-fat diet promoted OP development in vivo, and OX-LDL stimulated osteoblast apoptosis in vitro. GEN exhibited protective activities against OX-LDL-induced osteoblast apoptosis by increasing the NRF2 pathway and decreasing the NF-κB pathway. PDTC, an NF-κB inhibitor, could further promote the biological functions of GEN. In contrast, ML385, an NRF2 inhibitor, might eliminate GEN's protection. Conclusion: GEN suppressed OX-LDL-induced osteoblast apoptosis by regulating the NRF2/NF-κB signaling pathway. [ABSTRACT FROM AUTHOR]

The purpose of this paper is to study the impact of road structure on traffic noise from an urban network planning perspective, to provide solutions to control and prevent traffic noise at source. The process is divided into three parts. Firstly, the stochastic user equilibrium model for traffic flow allocation is improved to be more accurate. Secondly, a global traffic noise simulation method is proposed for the network based on the noise prediction model, and the noise data is visualized. Finally, a noise evaluation system is established to study the traffic noise evolution patterns under different road structures. The results show that with the increase of network density, the traffic noise value of the network decreases. The layout of one-way roads will reduce the regional traffic noise value. The impact of road grade adjustments on traffic noise is positively correlated with the rate of change in road capacity. The study can provide an effective basis for control traffic noise from the network planning stage. [ABSTRACT FROM AUTHOR]

• A road traffic noise evaluation method for heterogeneous crowds is presented. • Heterogeneous crowd distribution, age and acoustic requirements are considered. • A noise response curve (NRC) is established. • The application verified the reasonableness and practicality of the method. The primary objective of this study is to develop an urban traffic noise exposure evaluation method for differential crowd characteristics, which takes crowd distribution, ages and acoustic requirements into consideration. First, road traffic noise data are simulated using a line-source model, and crowd distribution data are obtained by the decomposition of demographic data based on the different building attraction coefficients of the heterogeneous crowd. Then, the noise response curve (NRC) of the heterogeneous crowd is established with the normalization of subjective feelings of respondents and calibrating the response level of each age group in combination. Finally, with the help of the proposed noise evaluation indicators, and considering the noise function requirements (NFR), the traffic noise evaluation of differential crowd characteristics is realized. The method is applied to a typical scenario, and the results and analysis verified the reasonability and practicability of the method in urban traffic noise exposure evaluation. [ABSTRACT FROM AUTHOR]

Speech emotion recognition is an important task with a wide range of applications. However, the progress of speech emotion recognition is limited by the lack of large, high-quality labeled speech datasets, due to the high annotation cost and the inherent ambiguity in emotion labels. The recent emergence of large-scale video data makes it possible to obtain massive, though unlabeled speech data. To exploit this unlabeled data, previous works have explored semi-supervised learning methods on various tasks. However, noisy pseudo-labels remain a challenge for these methods. In this work, to alleviate the above issue, we propose a new architecture that combines cross-modal knowledge transfer from visual to audio modality into our semi-supervised learning method with consistency regularization. We posit that introducing visual emotional knowledge by the cross-modal transfer method can increase the diversity and accuracy of pseudo-labels and improve the robustness of the model. To combine knowledge from cross-modal transfer and semi-supervised learning, we design two fusion algorithms, i.e. weighted fusion and consistent & random. Our experiments on CH-SIMS and IEMOCAP datasets show that our method can effectively use additional unlabeled audio-visual data to outperform state-of-the-art results. [ABSTRACT FROM AUTHOR]

Patterned media technology is a promising approach for future high density magnetic data storage. The bit-to-bit spacing becomes smaller as recording density increases. As a result, the magneto-static field (Hd) from the surrounding bits becomes stronger, and contributes more to the broadening of the switching field distribution (SFD) of targeted bit. This paper reports the investigations between possible bit-cell arrangements and SFD broadening caused by the magneto-static field. The possible bit-cell arrangements are categorized into three groups—square arrangement, isosceles triangular arrangement, and equilateral triangular arrangement. The influences of manufacturing induced deviations, including bit-cell position offset and bit-cell size variation, over the Hd are also simulated. The results suggested that isosceles triangular arrangement can reduce the SFD broadening, without increasing the requirement on track positioning accuracy. [ABSTRACT FROM AUTHOR]

This paper explores the shingled writing head designs for making 10 Tb/in^2 patterned media data storage more feasible. The relationship between the writing ability improvement of the shingled writing head and the tip width is discussed. The results reveal that a small rounded corner design would be beneficial for improving the writing performance of the writing head. A corner gap, which is slightly smaller than the shield gap, will give the best writing performance. [ABSTRACT FROM AUTHOR]

• Anatomical prior information is introduced to estimate high-resolution CT images. • A united framework is applied instead of designing multiple models for individual human body sites. • An improved performance is achieved by Wasserstein GAN with an attribute loss function. Currently, many deep learning (DL)-based low-dose CT image postprocessing technologies fail to consider the anatomical differences in training data among different human body sites, such as the cranium, lung and pelvis. In addition, we can observe evident anatomical similarities at the same site among individuals. However, these anatomical differences and similarities are ignored in the current DL-based methods during the network training process. In this paper, we propose a deep network trained by introducing anatomical site labels, termed attributes for training data. Then, the network can adaptively learn to obtain the optimal weight for each anatomical site. By doing so, the proposed network can take full advantage of anatomical prior information to estimate high-resolution CT images. Furthermore, we employ a Wasserstein generative adversarial network (WGAN) augmented with attributes to preserve more structural details. Compared with the traditional networks that do not consider the anatomical prior and whose weights are consequently the same for each anatomical site, the proposed network achieves better performance by adaptively adjusting to the anatomical prior information. [ABSTRACT FROM AUTHOR]

Background: This study was undertaken to establish a rat bipedal walking model of cervical kyphosis (CK) associated with chronic forward flexed neck and assess the effects of chronic forward flexed neck on endplate chondrocytes. Methods: Forty-eight 1-month-old Sprague-Dawley rats were randomly divided into 3 groups: forward flexed neck group (n = 16), bipedal group (n = 16), and normal group (n = 16). Cervical curves were analyzed on a lateral cervical spine X-ray using Harrison's posterior tangent method before the experiment and at 2-week intervals for a 6-week period. Histologic changes in cartilaginous endplate chondrocytes were observed using hematoxylin and eosin (H&E) staining, transmission electron microscopy (TEM), and terminal deoxyribonucleotidyl transferase (TdT)-mediated dUTP nick-end labeling. Results: Radiographic findings suggested a significantly decreased cervical physiological curvature in the forward flexed neck group over the 6-week follow-up; normal cervical curves were maintained in other groups. The average cervical curvature (C2–C7) was − 7.6 ± 0.9° in the forward flexed neck group before the experiment, − 3.9 ± 0.8° at 2 weeks post-experiment, 10.7 ± 1.0° at 4 weeks post-experiment, and 20.5 ± 2.1° at the last follow-up post-experiment. Histologically, results of H&E staining unveiled that cartilaginous endplate chondrocytes were arranged in an irregular fashion, with the decreased number at the observation period; the incidence of apoptotic cells in the forward flexed neck group was noticeably higher at the 6-week follow-up than that in other groups. Conclusions: CK developed as the result of chronic forward flexed neck. Histologic changes suggested that chondrocyte apoptosis may play a critical role in the development of cervical kyphotic deformity associated with chronic forward flexed neck. [ABSTRACT FROM AUTHOR]

Nonlinear transition shift (NLTS) produced by the interaction between transitions is a major source of distortions in magnetic recording. NLTS must be reduced to realize higher density and maintain signal quality. However, there is some confusion about the production and behavior of NLTS in previous reports. In this paper, combined with micromagnetic simulation, we have developed a complete theoretical model to predict NLTS on perpendicular magnetic media. The results show that one-level NLTS (caused by one previous transition) will monotonically decrease as the present transition becomes further away from the previous one, and multilevel NLTS (caused by consecutive previous transitions) will firstly increase and quickly approach a constant value as there are more consecutive transitions. Particularly, it is demonstrated that NLTS and its impact on the media signal-to-noise ratio can be significantly limited by pattern constraints. • A complete theoretical model to predict nonlinear transition shift. • Micromagnetic simulation to verify the theoretical model. • Understanding of the behavior of nonlinear transition shift for complex patterns. • Nonlinear transition shift can be significantly limited by pattern constraints. [ABSTRACT FROM AUTHOR]

Multimodal sentiment analysis is an active research field that aims to recognize the user's sentiment information from multimodal data. The primary challenge in this field is to develop a high-quality fusion framework that effectively addresses the heterogeneity among different modalities. However, prior research has primarily concentrated on intermodal interactions while neglecting the semantic sentiment information conveyed by words in the text modality. In this paper, we propose the Sentiment Knowledge Enhanced Attention Fusion Network (SKEAFN), a novel end-to-end fusion network that enhances multimodal fusion by incorporating additional sentiment knowledge representations from an external knowledge base. Firstly, we construct an external knowledge enhancement module to acquire additional representations for the text modality. Then, we design a text-guided interaction module that facilitates the interaction between text and the visual/acoustic modality. Finally, we propose a feature-wised attention fusion module that achieves multimodal fusion by dynamically adjusting the weights of the additional and each modality's representations. We evaluate our method on three challenging multimodal sentiment analysis datasets: CMU-MOSI, CMU-MOSEI, and Twitter2019. The experiment results demonstrate that our model significantly outperforms the state-of-the-art models. The source code is publicly available at https://github.com/doubibobo/SKEAFN. • Propose a sentiment knowledge-enhanced attention fusion network (SKEAFN). • Build an additional knowledge graph to leverage explicit sentiment information. • Use a multi-head attention mechanism to model the interactions among modalities. • Develop feature-wised attention to adjust the contributions of multiple modalities. [ABSTRACT FROM AUTHOR]

Osteoporosis, characterized by decreased mineral density and bone mass, is triggered by various detrimental factors and often causes further complications, including fractures. Aberrant expression of microRNAs (miRs) has been associated with the pathogenesis of osteoporosis. Recently, miR-142 was reported to be downregulated in osteoblasts; however, the underlying mechanism of miR-142 in mediating the development of osteoporosis remains unclear. In the present study, high glucose induced the downregulation of miR-142 mRNA expression and promoted the apoptosis of MC3T3-E1 cells. miR-142-mimics significantly protected against high glucose-induced apoptosis, upregulated the expression levels of B-cell lymphoma 2 (Bcl-2) and downregulated the protein expression levels of β-catenin, Bcl-2 associated X (Bax) and caspase-3. Furthermore, β-catenin was identified as a direct target of miR-142 using luciferase reporter assays. Similar to the effects of miR-142 inhibitors, overexpression of β-catenin aggravated the apoptosis of MC3T3-E1 cells, as demonstrated by the upregulation of Bax and caspase-3, and the downregulation of Bcl-2 expression levels. In conclusion, miR-142 protects MC3T3-E1 cells against high glucose-induced apoptosis by targeting β-catenin. [ABSTRACT FROM AUTHOR]

Various schemes have been investigated to study the modeling of recording media, the writing/reading process, and the information recovery in magnetic recording systems, such as two-dimensional magnetic recording (TDMR) and bit-patterned media recording (BPMR). As the bit scales down to the size of the media particles, the channel modeling and simulation become more complicated because of the increasing media noise. Meanwhile, it relies on more powerful 2-D signal-processing algorithms error control codes. The contributions of this article are three-fold: 1) we improve the read process in terms of our proposed reliability for read response that captures the effects of inter-track interference (ITI), inter-symbol interference (ISI), and the randomness of grains; 2) we present a neural network block predictor (NNBP) and a logistic regression method to provide heuristic methods in studying data recovery; and 3) we investigate the most possible harmful transitions and patterns. Results show the NNBP achieves good performance and is good at detecting those totally isolated bits that are thought to be bad patterns from an intuitive point of view. [ABSTRACT FROM AUTHOR]

To achieve a high track density, two-dimensional magnetic recording (TDMR) is combined with shingled magnetic recording (SMR). SMR makes it possible to record 1 bit on a few grains. However, the performance will be remarkably deteriorated by the increased media noise, the inter-track and inter-symbol interference (ITI and ISI). Therefore, the application of effective equalization and error control coding are required. In this paper, we investigate a simple block-based neural network equalizer (NNE) that mitigates the influence of ITI and ISI. We compare the equalization effects of the NNE and a conventional 2-D equalizer with low-density parity-check (LDPC) coding based on a random Voronoi grain media model. Simulation results show the proposed block-based NNE achieves better bit error rate performance than the conventional 2-D linear equalizer followed by the a posteriori probability (APP) detector and a sum-product (SP) decoder. In addition, we find the block-based NNE is sensitive to write errors. [ABSTRACT FROM AUTHOR]

Highlights • Propose a deep weight fusion method for emotion recognition. • Conduct the cross-modal noise modeling for the audio and video data. • Use the deep belief network for multi-modal emotion features. Abstract The multi-modal emotion recognition lacks the explicit mapping relation between emotion state and audio and image features, so extracting the effective emotion information from the audio/visual data is always a challenging issue. In addition, the modeling of noise and data redundancy is not solved well, so that the emotion recognition model is often confronted with the problem of low efficiency. The deep neural network (DNN) performs excellently in the aspects of feature extraction and highly non-linear feature fusion, and the cross-modal noise modeling has great potential in solving the data pollution and data redundancy. Inspired by these, our paper proposes a deep weighted fusion method for audio-visual emotion recognition. Firstly, we conduct the cross-modal noise modeling for the audio and video data, which eliminates most of the data pollution in the audio channel and the data redundancy in visual channel. The noise modeling is implemented by the voice activity detection(VAD), and the data redundancy in the visual data is solved through aligning the speech area both in audio and visual data. Then, we extract the audio emotion features and visual expression features via two feature extractors. The audio emotion feature extractor, audio-net, is a 2D CNN, which accepting the image-based Mel-spectrograms as input data. On the other hand, the facial expression feature extractor, visual-net, is a 3D CNN to which facial expression image sequence is feeded. To train the two convolutional neural networks on the small data set efficiently, we adopt the strategy of transfer learning. Next, we employ the deep belief network(DBN) for highly non-linear fusion of multi-modal emotion features. We train the feature extractors and the fusion network synchronously. And finally the emotion classification is obtained by the support vector machine using the output of the fusion network. With consideration of cross-modal feature fusion, denoising and redundancy removing, our fusion method show excellent performance on the selected data set. [ABSTRACT FROM AUTHOR]

All-optical switching (AOS) is an emerging recording technology by using femtosecond laser to manipulate the ultrafast magnetization reversal for future super high density magnetic storage. TbCo films possess a ferrimagnetic structure and large perpendicular anisotropy and, thus, become a potential material for AOS. In this paper, we simulate the ultrafast magnetization reversal domain of Co-sublattice in TbCo films under 100 fs pulse laser with a general theoretical framework, and our simulation results show that TbCo films irradiated with a higher power fluence laser have a larger reversal domain of Co-sublattice due to the higher peak temperature and the further critical isotherm of electrons. Moreover, high Co content can help Co-sublattice form a large reversal domain, but is unable to obtain the highly uniform reversal domain perhaps attributed to the insufficient anti-ferromagnetic interaction between Tb atoms and Co atoms. [ABSTRACT FROM AUTHOR]

MnGa films are the promising magnetic recording materials and spintronic materials owing to their intrinsic properties, such as large magnetic anisotropy, high coercivity, moderate magnetization, and high spin polarization. In this paper, MnGa films with high coercivity and low surface roughness have been successfully fabricated onto MgO substrates by magnetron sputtering and post-annealing. Moreover, the effects of post-annealing temperature ( T) on crystalline structure, surface morphology, and magnetic performances of MnGa films have also been investigated. It is found that the crystallization temperature for MnGa films is 400 C. With increasing T, the crystallization degree enhances and an in-plane texture is formed. The grain size and surface roughness of MnGa films increase slowly when T is below 500 C, but they exhibit a rapid rise when T is above 500 C. As T increases, the coercivity ( H) and remanence squareness ratio ( S) for MnGa films improve monotonically, whereas saturation magnetization ( M) increases firstly and then drops. The increases in H, S, and M with T are attributed to the grains' growth and the improvement of crystallinity, and the decrease of M at higher T possibly is due to the partial oxidation of Mn. [ABSTRACT FROM AUTHOR]

The thickness dependence of magnetization reversal and coercivity behavior for La–Co substituted strontium hexaferrite (Sr- M ) films was investigated. It is found that perpendicular anisotropy appears only when film thickness ( t ) is above 110 nm. With increasing t , perpendicular anisotropy energy ( K u⊥ ) increases gradually to its maximum of 1.76×10 6 erg/cm 3 at t =300 nm, but turns to decrease when t >300 nm. Moreover, when t >110 nm, those films exhibit domains pinning or Stoner–Wohlfarth reversal model, present large K u⊥ values and a rapid increase in H c⊥ . However, while t ≤110 nm, Sr- M films show nucleation model of magnetization reversal and perform low coercivity. The origin of the coercivity varying with thickness should be correlated with the grain size and preferred orientations in Sr- M films. [ABSTRACT FROM AUTHOR]

Shingled magnetic recording (SMR) with exchange-coupled composite media provides a feasible way by enhancing the writing capability and reducing the media switching field for achieving ultrahigh areal density of magnetic storage. However, the media switching field is dependent on the angle between the easy axis and the applied field. The adjacent track encroachment induced by this angular disturbance may happen, which could increase recording noise and lead to measurable erasure. This problem is highlighted in SMR system because of the enlarged stray field. A trapezoidal shingled writing head is modeled to explore the spatial field distribution and how the writing performance is affected by the angular dependence of media coercivity. The results show that a larger easy axis deviation and a smaller anisotropy energy density ratio of hard region will be helpful for reducing switching field to improve writing capability. The analysis also suggests that, if properly arranged, the adjacent bits can suffer minimized encroachment caused by variation of angularly induced switching field and attain a satisfactory bit error rate of as low as < 5 \times 10^- 4 . [ABSTRACT FROM AUTHOR]