Search

Existing work on Multimodal Sentiment Analysis (MSA) utilizes multimodal information for prediction yet unavoidably suffers from fitting the spurious correlations between multimodal features and sentiment labels. For example, if most videos with a blue background have positive labels in a dataset, the model will rely on such correlations for prediction, while ``blue background'' is not a sentiment-related feature. To address this problem, we define a general debiasing MSA task, which aims to enhance the Out-Of-Distribution (OOD) generalization ability of MSA models by reducing their reliance on spurious correlations. To this end, we propose a general debiasing framework based on Inverse Probability Weighting (IPW), which adaptively assigns small weights to the samples with larger bias i.e., the severer spurious correlations). The key to this debiasing framework is to estimate the bias of each sample, which is achieved by two steps: 1) disentangling the robust features and biased features in each modality, and 2) utilizing the biased features to estimate the bias. Finally, we employ IPW to reduce the effects of large-biased samples, facilitating robust feature learning for sentiment prediction. To examine the model's generalization ability, we keep the original testing sets on two benchmarks and additionally construct multiple unimodal and multimodal OOD testing sets. The empirical results demonstrate the superior generalization ability of our proposed framework. We have released the code and data to facilitate the reproduction.

Existing studies on multimodal sentiment analysis heavily rely on textual modality and unavoidably induce the spurious correlations between textual words and sentiment labels. This greatly hinders the model generalization ability. To address this problem, we define the task of out-of-distribution (OOD) multimodal sentiment analysis. This task aims to estimate and mitigate the bad effect of textual modality for strong OOD generalization. To this end, we embrace causal inference, which inspects the causal relationships via a causal graph. From the graph, we find that the spurious correlations are attributed to the direct effect of textual modality on the model prediction while the indirect one is more reliable by considering multimodal semantics. Inspired by this, we devise a model-agnostic counterfactual framework for multimodal sentiment analysis, which captures the direct effect of textual modality via an extra text model and estimates the indirect one by a multimodal model. During the inference, we first estimate the direct effect by the counterfactual inference, and then subtract it from the total effect of all modalities to obtain the indirect effect for reliable prediction. Extensive experiments show the superior effectiveness and generalization ability of our proposed framework.

Although acute promyelocytic leukaemia (APL) is a highly curable disease, challenges of early death (ED) and relapse still exist, and real-world data are scarce in the ATRA plus ATO era. A total of 1105 APL patients from 1990 to 2020 were enrolled and categorized into three treatment periods, namely ATRA, ATRA plus ATO, and risk-adapted therapy. The early death (ED) rate was 20.2%, 10.1%, and 7.0%, respectively, in three periods, while there was no significant decline in the 7-day death rate. Consistently, the overall survival (OS) and disease-free survival (DFS) of APL patients markedly improved over time. Despite the last two periods exhibiting similar DFS, the chemotherapy load was substantially lower in Period 3. Notably, leveraging older age and higher WBC count (especially50 × 10

Marine natural capital is an important component of natural capital yields goods and service flows benefiting the human being. The emergy analysis method allows one to account for mass, energy and money flows in an ecosystem, providing technical support for assessing its broader value regarding our economic dependence. Thus, we used this method to evaluate the natural capital of the Zhoushan archipelago sea area from 2011 to 2016 and proposed a formula to estimate the marine organism’s transformity. The average total emergy of our study area was 6.93 × 1022 sej and emdollar was about 9.20 billion yuan, which is equivalent to 9.3% of the average regional GDP of 98.5 billion during the same period. The Zhoushan archipelago sea area has high emergy density (ED) and low emergy self-sufficiency ratio (ESR), which shows low input–output efficiency for local use. In addition, the high purchased emergy (PR), high emergy exchange ratio (EER) and low renewable resources emergy ratio (%R) imply an increasing dependence on the outside social and economic inputs. Overall, Zhoushan sea area was in an early but steady state of development. The results can serve as a benchmark for policy making and implementation to achieve local sustainable development. As a tool for emergy-based sea area capital assessment, the model is of great significance for quantifying the ecosystem service value and accounting for marine/land natural capital value.

Automatic weeding is a hot research field of smart agriculture, which has many benefits such as achieving precise weed control, saving human cost, and avoiding damage on crops, etc. Recently, many researchers have focused on the research using the deep learning method, such as the convolutional neural network (CNN) and recurrent neural network (RNN) and have achieved decent outcomes related to the automatic weed detection. However, there are still generally problems of the projects such as weak robustness and excessive reliance on a large number of samples. To solve these problems, a recognition algorithm for automatic identification and weed removal was designed, and a soybean field weed detection and localization method based on the fusion of multiple image processing methods was proposed in this study. The images and video stream were obtained through the camera mounted on a mobile robot platform. Firstly, the soil background inside the image was segmented from the foreground (including the weeds and crops) by setting the threshold for a specific color space (hue). Then, three different methods including the area threshold method, template matching and saturation threshold method were used to classify the crops and weeds. Finally, based on a proposed innovative voting method, the three recognition methods were comprehensively weighed and fused to achieve more accurate recognition and localization results of the crops and weeds inside the image. Experimental validations were carried out using the samples obtained through the moving platform, and the experimental results showed that the average accuracy of the proposed weed detection algorithm was as high as 98.21%, while the recognition error was only 1.79%. Meanwhile, compared with each single method as the scale threshold, template matching and saturation threshold, the fused method based on the weighted voting has been able to raise the average accuracy by 5.71%. Even though the samples used in the validations were limited in covering different scenarios, the high recognition accuracy has proved the practicability of the proposed method. In addition, the robustness test that images with raindrop and shadow interference in the complex and unstructured agricultural scene was carried out, and satisfied results showed that above 90% of the plant were successfully detected, which verified the fine adaptability and robustness of the proposed method.

Reversible phase-change behaviors of Ge&ndash, Sb&ndash, Te based superlattices (GST-SL) were studied by ab initio molecular dynamics (AIMD) simulations based on three models containing Ge/Sb intermixing, namely the Petrov-mix, Ferro-mix, and Kooi-mix models. The flipping behavior of Sb atoms was found in all the three GST-SL models in the melting process. Among them the Kooi-mix model exhibited the best stability, and the analyses of bond length distribution and electron localization function provided a better explanation on the phase transition of GST-SL. Finally, we proposed a fast switching model for GST-SL based on Sb flipping.

Regarding the radiation relationship between complex objects involved in the space environment, in this paper we use discrete coordinate method to calculate and analyze the coupled radiation-conduction heat transfer between complex objects. We simulate the heat conduction process by use of the finite volume method, and the overall temperature field is finally obtained. By comparing the calculation results of the program and Fluent software, the correctness of the coupled radiation-conduction heat transfer program is verified. The result shows that the radiation heat transfer between the targets could affect the temperature field under the conditions of self-emission and ambient radiation. And it also shows that an appropriate reduction of discrete points can reduce the calculation time and will not affect the calculation accuracy.

International guidelines do not recommend routine systemic antibiotic prophylaxis in thyroidectomy patients as post-operative wound infections are rare, though many surgeons prefer to give preoperative antibiotics. However, the molecular-based short-term consequences of one dose of antibiotics on the human intestinal microbiota are rarely explored.

A newly proposed multi-directional forging (MDF) was successfully applied to a commercially pure titanium (CP Ti). Severe plastic deformation would result in significant and complex changes of microstructure and mechanical properties, so microstructure characterization and a mechanical test of CP Ti were conducted after different cycles of MDF. The results demonstrated that dynamic recrystallization was the dominant grain refinement mechanism of MDF CP Ti. With increasing the cycles of MDF, grain size, fraction of low angle grain boundaries and dislocations density decreased due to grain refined. After three cycles of MDF, the mean grain size was about 200 nm. The values of tensile strength and hardness increased significantly from zero cycles to one cycle of MDF, but increased slowly after one MDF cycle. Numerous dimples and tear ridges were present, but the dimples were smaller and shallower with increasing cycles of MDF.

Inductors are critical storage components in power converters. Their large size is, however, a major bottleneck for power module integration and efficient power management. High-density inductors can migrate the power converter close to the load. This can lead to lower losses and more efficient and granular power delivery. However, traditional magnetic materials cannot achieve high power density and current handling. Nanomagnetic films provide unique opportunities to enhance permeability, reduce core losses, and integrate inductors close to the chip. Introduction of magnetic anisotropy into the films provides an additional degree of freedom to further enhance current handling. This chapter reviews the fundamentals of nanomagnetic materials and designs for inductor cores with various topologies and illustrates their superiority over traditional materials. Such films can also be applied to RF components for improved inductance density.

The aim of this study is to reveal the fatigue failure mechanism of the fractured steel wheel after the cornering fatigue test. High power scanning electron microscopy is performed to analyse the fractured surfaces of the wheel and to examine the fatigue cracks in greater detail. The results show that the failure locations are mainly around the cooling hole and the bolt hole of the disc. The fatigue cracks are distributed along the circumference of the disc, and their lengths are approximately equal to each other. Further fractographic analysis reveals that multiple crack origins are detected on the surface of the disc, at the same time the fatigue crack features are very obvious. In addition, transgranular cleavage, ductile fatigue striations and brittle striations are also found. These failures are modes of fretting fatigue damage and consequence of stress concentration, which are attributed to the surface quality, the thickness variation and the corner radius of the disc.

A computational methodology is proposed to simulate wheel dynamic cornering fatigue test and estimate its’ multi-axial fatigue life. The technique is based on the critical plane theory and the finite element methods. The prediction of fatigue life is found to be in close agreement with the corresponding experiment. The stress states of wheel are basically biaxial tensile and compression normal stresses during the prototype test. The principal stresses are not proportional and the unstable principle plane is changing with loading direction, which indicates that the fatigue crack may occur first in the circumferential direction of steel wheel.

Since the traditional electrical system data monitor equipment has some shortage, such as limited monitoring objects, single monitoring direction, the authors combine computer science Internet technologies with electric power system production process and extend the scope of monitoring system. This paper analyzes the key technology by which the operator monitoring system can be realized, and puts forward the scheme, which based on.NET platform, of recording operator’s behaviors during electric power generation, and it is implemented in WINDOWS system.

Quantitative phase microscopy, as a non-destructive and non-invasive measurement technique, can indirectly reflect three-dimensional (3D) morphology and optical properties of transparent microstructure object by measuring phase information. In recent years, this kind of technique has been widely used to detect and investigate the characteristics of biological cells and it has become more and more important in the field of modern biomedical and life science. In this paper, only by using a single cube beamsplitter interferometer, a simple single-shot dual-channel quantitative phase microscopic measurement technique is demonstrated for 3D quantitative phase imaging of biological cells. In the proposed method, a conventional non-polarized cube beamsplitter is the most pivotal element. Unlike its traditional application method, the cube beamsplitter is tilted in a nonconventional configuration and the illumination beam is only incident on the left (or right) half of the cube beamsplitter (just the one side of central semi-reflecting layer), and a very small angle is introduced between the central semi-reflecting layer and the optical axis of incident beam. Based on the light splitting characteristic of the cube beamsplitter, two replicas of incident beam are generated. These two generated replicas (transmission beam and reflection beam) are of symmetry with respect to each other, and they will encounter and form interference when the direction of the incident beam meets a certain condition. Adjust the sample to a suitable position and make it only contact one half of incident beam, and the modulated beam will be seen as the object beam and the remaining clean half of incident beam as the reference beam. When the interference phenomenon occurs, two interference channels with a relative π (rad) phase-shift in one interferogram are acquired simultaneously only using one digital camera, and the higher spatial frequency of interference fringes can be achieved by adjusting a relatively big angle between the central semi-reflecting layer and the optical axis of incident beam. Because of the off-axis interference mode, we only need to record one interferogram to gain the continuous phase information and avoid using complex phase-shift techniques. At the same time, this proposed method is of simple structure and easy to operate due to using less ordinary off-the-shelf optical elements. All these simplify the structure of the system and reduce the cost of the system as much as possible. Finally, the phase information of paramecium is successfully obtained from different interference channels respectively. Furthermore, according to the characteristic of π (rad) phase-shift, we also realize the calibration and determination of ultimate precise phase information of sample by using the method of averaging between these two channels. The experimental results show that our proposed method is suitable for 3D surface morphology measurement of small transparent samples.

This paper presents materials modeling, design, processing, integration and characterization of a new class of nanomagnetic structures for coupling and shielding in wireless charging and power conversion applications. Wireless power transfer applications such as wireless charging, operating at 6.78 MHz, require high-performance magnetic materials for enhancing the coupling between transceiver and receiver coils as well as for suppressing electromagnetic interference (EMI) shielding. This research describes two novel magnetic structures for coupling inductors and ultra-thin EMI shields. A novel vertically aligned magnetic composite structure was demonstrated for the coupling inductor. This structure is shown to result in permeabilities of above 500 and loss tangent of 0.01, which enhances the coupling inductance by 3–5x at 6.78 MHz, and also enhances the power-transfer efficiency by 2x. The second part of this paper presents the modeling, design and fabrication of nanomagnetic structures for ultra-thin EMI shields in wireless power transfer applications. The ultra-thin EMI shields for wireless power transfer described in this research can achieve greater than 20dB attenuation at 6.78 MHz even for 3–5µm shield thickness.

Rationale: Necrosis is one of the main forms of cardiomyocyte death in heart disease. Recent studies have demonstrated that certain types of necrosis are regulated and programmed dependent on the activation of receptor-interacting serine/threonine-protein kinase (RIPK) 1 and 3 which may be negatively regulated by Fas-associated protein with death domain (FADD). In addition, microRNAs and long noncoding RNAs have been shown to play important roles in various biological processes recently. Objective: The purpose of this study was to test the hypothesis that microRNA-103/107 and H19 can participate in the regulation of RIPK1- and RIPK3-dependent necrosis in fetal cardiomyocyte-derived H9c2 cells and myocardial infarction through targeting FADD. Methods and Results: Our results show that FADD participates in H 2 O 2 -induced necrosis by influencing the formation of RIPK1 and RIPK3 complexes in H9c2 cells. We further demonstrate that miR-103/107 target FADD directly. Knockdown of miR-103/107 antagonizes necrosis in the cellular model and also myocardial infarction in a mouse ischemia/reperfusion model. The miR-103/107-FADD pathway does not participate in tumor necrosis factor-α–induced necrosis. In exploring the molecular mechanism by which miR-103/107 are regulated, we show that long noncoding RNA H19 directly binds to miR-103/107 and regulates FADD expression and necrosis. Conclusions: Our results reveal a novel myocardial necrosis regulation model, which is composed of H19, miR-103/107, and FADD. Modulation of their levels may provide a new approach for preventing myocardial necrosis.

Cell reprogramming is a process involved in changing epigenetic landscapes, including histone modification, DNA methylation, and expression of noncoding RNAs; and reprogramming finally leads to changes in gene expression profile and cell fate. A great challenge to this field is to overcome epigenetic suppression exerted by highly differentiated cells of those key regions that are critical for establishment and maintenance of final cell types or induced pluripotent stem cells (iPSCs). As a new class of small non-coding RNAs, piwi-interacting RNAs (piRNAs) have been shown to play important roles in transposon silencing, transcriptional/post-transcriptional regulation, and epigenetic modifications. In this review, we discuss recent advances in which piRNAs were proposed or shown to be barriers to reprogramming suppression through epigenetic silencing, and it may be necessary to overcome this piRNA-derived barrier to achieve final cellular status during reprogramming. Therefore, gaining deeper insights into the mechanism(s) by which piRNAs mediate epigenetic regulation of gene expression, genome stability and chromatin status may offer a new avenue for efficient reprogramming of somatic cells toward a pluripotent state.

In this paper, estimation of channels with large delay spread but with few nonzero taps, such as those encountered in hilly broadcast wireless communications, are considered. Exploiting the sparsity, a channel estimate can be obtained by using a matching pursuit (MP) algorithm. To improve the performance of MP algorithm based estimation, the orthogonal matching pursuit (OMP) algorithm for channel estimation is proposed. In OMP, the re-selection problem of MP algorithm is avoided by using the stored dictionary at each iteration, and faster convergence to a sparse solution is obtained. we proposed to use MP algorithm based on Pseudorandom Sequences for training sequences for channel estimation. Using the proposed method, the main taps distorted by the projection of other taps is eliminated by the dictionary with orthogonal property, and more accurate channel estimates can be obtained. The results of channel estimates by using MP, OMP and the proposed method are compared, verifying that the proposed method outperforms the MP and OMP methods, with the same computational complexity as MP algorithm.

AIS (automatic identification system), a new type of the radio navigation system at sea, can provide more target information than radar. It is very important how to make a desk officer intuitively observe the information so as to conveniently judge the motion tendency of the targets around own ship. We had developed a system of the AIS information display based on the platform of ENC (electronic nautical chart) by using Microsoft VC++ software tool. The 6 bit encapsulated data from the AIS output interface were received and decoded. After being converted into ASCII code in the format, the AIS information was showed on the platform of ENC with both graphic and alphanumeric mode, complied with the international criterion. The results of a spot test show that the system has a lot of advantages including good interface and so on.

By using projection augmented plane wave method (PAW) and based on the density functional theory, the stability of lattice dynamics and the magnetism of ordered crystalline alloy -Fe4N are studied at high external pressures. In comparison with the phonon spectrum of -Fe4N without considering the spin-polarization, it is found that the ground-state lattice dynamics stability of the ferromagnetic phase -Fe4N is induced by the spontaneous magnetization at pressures below 1 GPa. The phonon spectra at point (0.37, 0.37, 0) in line , points X and M become softening at pressures between 1.03 and 31.5 GPa. The pressure-induced effect and the spontaneous magnetization effect on the atoms reach a stable equilibrium state at the pressures between 31.5 and 60.8 GPa, which result in the phonon spectrum stability. As the pressure exceeds 61.3 GPa, the system becomes more instable dynamically with the increase of the external pressure. The softening at point M of the acoustic phonon is treated by the soft-mode phase theory at 10 GPa, and a new dynamic stability high-pressure phase with a space group of P2/m is found. This new phase is thermodynamically stable and possesses the same magnetic moments as that of -Fe4N at pressures below 1 GPa. The enthalpy value of the phase P2/m is less than that of phase at the pressures between 2.9 and 19 GPa, therefore its ground-state structure is more stable. As the pressure exceeds 20 GPa, both phases possess almost the same magnetic moments.

The ordered crystalline Fe3Pt invar alloy is in a special magnetic critical state under which the lattice dynamic stability of the system is extremely sensitive to the external pressure. Using projection augmented plane wave method based on the density functional theory, we calculate the dependences of enthalpy and magnetism of different crystalline alloys of Fe3Pt on external pressure. The results reveal that the P4/mbm structure is thermodynamically stable under pressures below 18.54 GPa. The total magnetic moments of Pm3m, I4/mmm and DO22 structures decrease rapidly and oscillate near the ferromagnetic collapse critical pressure. Furthermore in I4/mmm and DO22 structures the atomic magnetic moment of Fe1 reverses near the critical pressure. At a pressure of 43 GPa, the micro-ferrimagnetic property in DO22 structure becomes apparently strengthened while its volume increases rapidly. The lattice dynamic calculation shows that the spontaneous magnetization of the system in ferromagnetic states induces the softening of the transverse acoustic phonon TA1(M) in the Fe3Pt of Pm3m structure, and there exists a strong spontaneous volume magnetostriction at pressures below 26.95 GPa. Especially, in the pressure range between the ferromagnetic collapse critical pressure 41.9 GPa and the magnetism completely disappearing pressure 57.25 GPa, the lattice dynamic stability is more sensitive to the pressure. The pressure induces the stability of the phonon spectrum of the system at pressures above 57.25 GPa.