Search

With the powerful learning ability of deep convolutional networks, deep clustering methods can extract the most discriminative information from individual data and produce more satisfactory clustering results. However, existing deep clustering methods usually ignore the relationship between the data. Fortunately, the graph convolutional network can handle such relationship, opening up a new research direction for deep clustering. In this paper, we propose a cross-attention based deep clustering framework, named Cross-Attention Fusion based Enhanced Graph Convolutional Network (CaEGCN), which contains four main modules: the cross-attention fusion module which innovatively concatenates the Content Auto-encoder module (CAE) relating to the individual data and Graph Convolutional Auto-encoder module (GAE) relating to the relationship between the data in a layer-by-layer manner, and the self-supervised model that highlights the discriminative information for clustering tasks. While the cross-attention fusion module fuses two kinds of heterogeneous representation, the CAE module supplements the content information for the GAE module, which avoids the over-smoothing problem of GCN. In the GAE module, two novel loss functions are proposed that reconstruct the content and relationship between the data, respectively. Finally, the self-supervised module constrains the distributions of the middle layer representations of CAE and GAE to be consistent. Experimental results on different types of datasets prove the superiority and robustness of the proposed CaEGCN.

In this paper, we consider the problem of descriptors construction for the task of content-based image retrieval using deep neural networks. The idea of neural codes, based on fully connected layers activations, is extended by incorporating the information contained in convolutional layers. It is known that the total number of neurons in the convolutional part of the network is large and the majority of them have little influence on the final classification decision. Therefore, in the paper we propose a novel algorithm that allows us to extract the most significant neuron activations and utilize this information to construct effective descriptors. The descriptors consisting of values taken from both the fully connected and convolutional layers perfectly represent the whole image content. The images retrieved using these descriptors match semantically very well to the query image, and also they are similar in other secondary image characteristics, like background, textures or color distribution. These features of the proposed descriptors are verified experimentally based on the IMAGENET1M dataset using the VGG16 neural network., Comment: 8

Deep reinforcement learning (DRL) has been adopted to provide a more powerful content caching policy. The current DRL-based scheme considers the requests for the same size and updates the caching for each request. However, the real-world data delivery systems usually refresh the content cache periodically, with different sizes of requests. To satisfy the real-world requirements, this study proposes a novel size adaptive content caching algorithm using DRL, termed SACC. SACC models the requests with random sizes and updates the cache after a batch of requests. Technically, SACC utilizes the Actor-Critic framework, which is able to process large discrete action space. SACC comprehensively considers the short-, medium- and long-term requests as the state to train the actor network. The reward is modeled as the cache hit rate. Once an action is selected from the policy network, it is expended to its k nearest neighbors. The critic network finds the action with the best reward from the k actions. The performance of the proposed SACC is evaluated through computer simulation. The experimental results showed that SACC could train the network much more efficiently and improve the cache hit rate by as much as 4% when comparing to the state-of-art DRL-based scheme

The rapid growth of information technology and Internet applications has provided users with an explosion of information. Mobile e-commerce applications and web search engines are of great interest in extracting representative information from the original abundant information. However, the information extracted by several existing methods, such as top-k, are often quite similar, which is difficult to meet users’ demand for diversified information. In order to increase the diversity of representative information, this article proposes a data-driven approach to automatically identifying a subset of the original dataset that can cover more themes and content. The data-driven approach consists of two stages. First, a new unified similarity measure is proposed for handling dataset with categorical and numeric attributes. We inject external knowledge and attribute interactions into the similarity learning process to improve the accuracy of similarity estimation between data objects. Second, we develop an enhanced density peaks clustering algorithm based on shared nearest neighbors to automatically identify representative objects according to the previous estimated similarity. The enhanced density peaks algorithm takes the local structure in the entire data space into consideration, which makes the proposed approach relatively insensitive to variations in dataset’ density and dimensionality. Theoretical analysis demonstrates that the time complexity of the proposed approach can achieve the best $\bm {O}({\bm {N}\log \bm {N}})$ . Extensive comparison experiments were conducted on artificial and real-world datasets. The experimental results demonstrate the effectiveness and robustness of the proposed approach.

β-titanium shape memory alloys are used as shape memory-based biomedical applications and superelastic materials because they are made up of non-toxic components. Also, their behaviors are more semilar with that of human bone and dental biological tissues, such as biocompatibility, superior corrosion resistance, low density, and low modulus of elasticity. In the current study, the effect of Ta additive on the physical properties of Ti-Nb alloy has been investigated. Ti-Nb-Ta alloy samples were produced such that the Tantalum (Ta) element was substituted instead of Niobium (Nb), and the induction arc melting was used to melting the mixed powders. To investigate the crystal structure, microstructure, and phase transformation temperatures of all samples, the X-Ray Diffraction (XRD), Scanning Electron Microscopic (SEM), and Diffraction Scanning Calorimetry (DSC) measurements have been used. Both β-rich and α peaks were observed in the XRD and SEM measurements. Also, the DSC results revealed that the Ti-Nb-Ta alloys have only austenite transformation temperatures and their phase austenite phase transformation were decreased by the effect of Ta additive, additionally, the enthalpy change in〖 E〗_3 sample ( Ti_75.5 Nb_22.5 Ta_2 (%at)) has the highest value, while E_4 sample ( Ti_75.5 Nb_21.5 Ta_3 (%at)) has the lowest respective value.

Fuzzy c-means (FCM) algorithms with spatial information have been widely applied in the field of image segmentation. However, most of them suffer from two challenges. One is that introduction of fixed or adaptive single neighboring information with narrow receptive field limits contextual constraints leading to clutter segmentations. The other is that incorporation of superpixels with wide receptive field enlarges spatial coherency leading to block effects. To address these challenges, we propose fuzzy Students t-distribution model based on richer spatial combination (FRSC) for image segmentation. In this Paper, we make two significant contributions. The first is that both narrow and wide receptive fields are integrated into the objective function of FRSC, which is convenient to mine image features and distinguish local difference. The second is that the rich spatial combination under Students t-distribution ensures that spatial information is introduced into the updated parameters of FRSC,which is helpful in finding a balance between the noise-immunity and detail-preservation. Experimental results on synthetic and publicly available images, further demonstrate that the proposed FRSC addresses successfully the limitations of FCM algorithms with spatial information and provides better segmentation results than state-of-the-art clustering algorithms. National Natural Science Foundation of China (Grant Number: Grant 61871259, Grant 61861024); Natural Science Basic Research Program of Shaanxi (Grant Number: 2021JC-47); Key Research and Development Program of Shaanxi (Grant Number: 2021ZDLGY08-07).

© 2021 Thomas Telford Ltd.Binary granular soil mixtures, as common heterogeneous soils, are ubiquitous in nature and man-made deposits. Fines content and particle size ratio are two important gradation parameters for a binary mixture, which have potential influences on mechanical behaviours. However, experimental studies on drained shear behaviour considering the whole range of fines content and different particle size ratios are scarce in the literature. For this purpose, a series of drained triaxial compression tests was performed on dense binary silica sand mixtures with four different particle size ratios to investigate systematically the effects of fines content and particle size ratio on the drained shear behaviours. Based on these tests, the strength-dilation behaviour and critical state behaviour were examined. It was observed that both fines content and particle size ratio have significant influence on the stress-strain response, the critical state void ratio, the critical state friction angle, the maximum dilation angle, the peak friction angle and the stress-dilatancy relation. The underlying mechanism for the effects of fines content and particle size ratio was discussed from the perspective of the kinematic movements at particle level.

Problem definition: We analyze a firm that sells repeatedly to a customer population over multiple periods. Although this setting has been studied extensively in the context of dynamic pricing—selling the same product in each period at a varying price—we consider intertemporal content variation, wherein the price is the same in every period, but the firm varies the content available over time. Customers learn their utility on purchasing and decide whether to purchase again in subsequent periods. The firm faces a budget for the total amount of content available during a finite planning horizon, and allocates content to maximize revenue. Academic/practical relevance: A number of new business models, including video streaming services and curated subscription boxes, face the situation we model. Our results show how such firms can use content variation to increase their revenues. Methodology: We employ an analytical model in which customers decide to purchase in multiple successive periods and a firm determines a content allocation policy to maximize revenue. Results: Using a lower bound approximation to the problem for a horizon of general length T, we show that, although the optimal allocation policy is not, in general, constant over time, it is monotone: content value increases over time if customer heterogeneity is low and decreases otherwise. We demonstrate that the optimal policy for this lower bound problem is either optimal or very close to optimal for the general T period problem. Furthermore, for the case of T = 2 periods, we show how two critical factors—the fraction of “new” versus “repeat” customers in the population and the size of the content budget—affect the optimal allocation policy and the importance of varying content value over time. Managerial implications: We show how firms that sell at a fixed price over multiple periods can vary content value over time to increase revenues.

The synergistic effect of Mo and Zr additions on microstructure evolution, room-temperature fracture toughness and microhardness of Nb-22Ti-15Si-xMo-yZr (x = 4,8, y = 3,6) alloys manufactured by laser directed energy deposited (L-DED) have been investigated. The major phases in as-deposited 4Mo-3Zr alloy are Nb solid solution (Nbss), Nb3Si and γ-Nb5Si3. The Nb3Si phases disappear with increasing Mo content to 8 at.% or increasing Zr content to 6%. γ-Nb5Si3 precipitates formed in the Nbss of as-deposited xMo-yZr alloys due to the cyclic re-heating, and the γ-Nb5Si3 precipitates and Nbss matrix exhibits orientation relationship (OR) of [001]Nbss//[ 1 ¯ 112]γ and (110)Nbss//(01 1 ¯ 0)γ in as-deposited 4Mo-6Zr alloy. The microstructure of xMo-yZr alloys became relative homogeneity and the Nbss matrix showed the better continuity after the heat treatment (1400 °C/30 h). The Nb3Si phase has transformed into Nbss and α-Nb5Si3 with the OR have been determined as {100}Nbss//{100}α and{111}Nbss//{111}α in the heat-treated 4Mo-3Zr alloy. When Mo content increase to 8at.%, a part of γ-Nb5Si3 phase also transformed into the α-Nb5Si3 phase with the OR have been determined as {11 2 ¯ 0}γ//{110}α and {10 1 ¯ 0}γ//{111}α after heat treatment. Among the as-deposited alloys, the 4Mo-6Zr alloy has the highest fracture toughness KQ with the lowest hardness. The average KQ of 4Mo-6Zr and 8Mo-6Zr alloys increased by 28% and 30% after the heat treatment, and reached 13.61 and 11.43 MPa•m1/2, respectively. The hardness of Nb-22Ti-15Si-xMo-yZr alloys is significantly decreased after the heat treatment.

We propose the option realized variance as an observable variable to summarize the information from high-frequency option data. This variable aggregates intraday option returns from midquote prices to compute an option’s total variability for a given day, providing additional information about the jump activity in the data generating process. Using the S&P 500 index time series and options data, this paper documents the performance of this realized measure in predicting the index realized variance as well as the equity and variance risk premiums. We estimate an option pricing model and analyze its parameter estimates. Our results show that excluding high-frequency option information produces significant differences in variance jump parameters, estimated risk premiums, and option pricing errors. This paper was accepted by Tyler Shumway, finance.

A detailed description is given of how the liquid water content (LWC) and the ice water content (IWC) can be determined accurately and absolutely from the measured water Raman spectra of clouds. All instrumental and spectroscopic parameters that affect the accuracy of the water-content measurement are discussed and quantified; specifically, these are the effective absolute differential Raman backscattering cross section of water vapor , and the molecular Raman backscattering efficiencies ηliq and ηice of liquid and frozen microparticles, respectively. The latter two are determined following rigorous theoretical approaches combined with Raman Lidar for Atmospheric Moisture Sensing (RAMSES) measurements. For ηice, this includes a new experimental method that assumes continuity of the number of water molecules across the vertical extent of the melting layer. Examples of water-content measurements are presented, including supercooled liquid-water clouds and melting layers. Error sources are discussed; one effect that stands out is interfering fluorescence by aerosols. Aerosol effects and calibration issues are the main reasons why spectral Raman measurements are required for quantitative measurements of LWC and IWC. The presented study lays the foundation for cloud microphysical investigations and for the evaluation of cloud models or the cloud data products of other instruments. As a first application, IWC retrieval methods are evaluated that are based on either lidar extinction or radar reflectivity measurements. While the lidar-based retrievals show unsatisfactory agreement with the RAMSES IWC measurements, the radar-based IWC retrieval which is used in the Cloudnet project performs reasonably well. On average, retrieved IWC agrees within 20% to 30% (dry bias) with measured IWC.

The synergy of active and passive microwave measurements has attracted considerable attention in recent years since they offer complementary information on the characteristics of the observed target (e.g., soil moisture), which motivates the launch of NASA's Soil Moisture Active Passive (SMAP) mission. An assumption of a near-linear relationship between active and passive measurements has been made in the SMAP active-passive baseline algorithm, which is essential to downscale coarse-resolution radiometer brightness temperature (TB) using high-resolution radar backscatter (σ⁰) but has not yet been fully tested under a wide range of ground conditions. Motivated by this, we first examined the validity of the linear assumption by using concurrent and coincident SMAP active and passive observations under diverse environmental factors (e.g., land cover, climate types, terrain and its complexity, soil texture, vegetation coverage, soil moisture, and its dynamics). We also adopted SMAP enhanced TB to evaluate the performance of the disaggregated TB at the same grid resolution of 9 km. The results reveal there is a generally good linear relationship between σ⁰ (no matter in dB or in linear unit) and TB at a global scale. There is no significant difference in the correlation among the four polarization combinations ( $σ ⁰_{hh}$ versus TB $_{h}$ , $σ⁰_{hh}$ versus TB $_{v}$ , $σ⁰_{vv}$ versus TB $_{h}$ , and $σ⁰_{vv}$ versus TB $_{v}$ ) with the $σ⁰_{vv}$ and TB $_{h}$ combination displaying an overall slightly higher correlation. The linear relationship between σ⁰ and TB is significantly affected by environmental factors. Particularly in bare soils and densely vegetated areas (e.g., large forest fraction and vegetation coverage), and arid and polar climate zones, the linear correlation between active and passive measurements worsens, whereas it is favorable in moderate vegetation and soil moisture as well as large soil moisture dynamic conditions. Interestingly, the linear correlation generally decreases as sand content increases while increases with the increase of clay content. The absolute linear correlation coefficient is higher with larger soil moisture dynamics. When compared to SMAP enhanced TB, it shows the linear assumption may have more influence on the correlation (i.e., temporal evolution) of downscaled TB than its absolute accuracy. These findings can enhance the understanding of the geophysical relationship between radar and radiometer signatures, and thus benefit active-passive joint algorithms for future satellite missions.

With the introduction of artificial intelligence (AI), current deep learning-driven change detection methods generally regard changes as a type of specific land cover information and try to detect them simply by using existing semantic labeling models. In practice, the changes that occur on different land cover categories may appear completely different for remote sensing images, making it difficult to detect multiple categories of changes with an end-to-end model. In addition, training such networks requires a large amount of prelabeled references, which are labor-intensive and time-consuming. Motivated by this observation, we integrate dual learning algorithm and disentangled representation theory to develop a novel approach, named content-invariant dual learning (CiDL), for either supervised or unsupervised change detection in remote sensing images. In our framework, two opposite Y-shaped networks, each of which consists of two encoders and one decoder, are introduced to translate bitemporal images from their original domains to each others, where their intrinsic content features are retained, while their style features are consistent. By training our hybrid framework, even without references, this method learns a category-wise cross-domain translation to suppress the discrepancies in paired unchanged regions, and meanwhile highlight those in paired changed regions. The experimental results on two typical change detection data sets and the comparison with other state-of-the-art deep learning-driven methods verify the effectiveness and competitiveness of our proposed CiDL.

Unsupervised Domain adaptation (UDA) attempts to boost the performance on an unlabeled target domain by transferring knowledge from a labeled source domain. The previous models consider the domain-level discrepancy while neglecting the content-level distinction. To further decrease the distribution gap between different domains, this paper proposes a novel UDA approach with content-wise alignment for hyperspectral image classification (HSIC). We accomplish feature alignment with content-wise discrepancy reduction through an adversarial framework for the first time. Expressly, the core of the proposed content-wise scheme is integrated with a class-level and a style-perceive-level regularized alignment to strengthen the representation of the invariant feature. The experimental analysis demonstrates that the proposed model achieves more effective performance than other domain adaptation methods for HSI.

BACKGROUND Dietary guidelines recommend limiting the intake of added sugars. However, despite the public health importance, most countries have not mandated the labeling of added sugar content on packaged foods and beverages, making it difficult for consumers to avoid products with added sugar, and limiting the ability of policymakers to identify priority products for intervention. OBJECTIVE To develop a machine learning approach for the prediction of added sugar content in packaged products using available nutrient, ingredient, and food category information. DESIGN The added sugar prediction algorithm was developed using k-Nearest Neighbors (KNN) and packaged food information from the US Label Insight dataset (n = 70,522). A synthetic dataset of Australian packaged products (n = 500) was used to assess validity and generalization. Performance metrics included the coefficient of determination (R2), mean absolute error (MAE), and Spearman rank correlation (ρ). To benchmark the KNN approach, the KNN approach was compared to an existing added sugar prediction approach that relies on a series of manual steps. RESULTS Compared to the existing added sugar prediction approach, the KNN approach was similarly apt at explaining variation in added sugar content (R2 = 0.96 vs. 0.97 respectively) and ranking products from highest to lowest in added sugar content (ρ = 0.91 vs. 0.93 respectively), while less apt at minimizing absolute deviations between predicted and true values (MAE = 1.68 g vs. 1.26 g per 100 g or 100 mL respectively). CONCLUSIONS KNN can be used to predict added sugar content in packaged products with a high degree of validity. Being automated, KNN can easily be applied to large datasets. Such predicted added sugar levels can be used to monitor the food supply and inform interventions aimed at reducing added sugar intake.

A thesis for interaction is proposed as a foundation for theory of interaction. It formulates the minimal content computability and the maximal channel computability of all models of interaction by fixing a minimal model and a maximal model. It is shown that the interactive extensions of the recursive function model, the Turing machine model and the λ-calculus model all fall within the realm of the thesis. A major technical contribution is the design of interpreters in the maximal model for the models that fit in the thesis.

In microwave remote sensing applications, some areas need to be observed frequently. Compared with the historical observations, the current observation has a little new content. This means that the increment is sparse, making it possible for imaging with under sampled echoes. For the increment imaging, we first establish the linear observation model and the increment model, making it a two-variable optimization problem. Then, the block coordinate descent strategy is used to implement the increment imaging. To estimate the unknown phase matrix, we make the uniform quantification assumption, and an extended Lagrangian multiplier method is applied. For the sparse increment, Bayesian compressed sensing (BCS) is used. Two different resolution scenes are used to verify the feasibility of this algorithm. The performance under different signal-to-noise ratios (SNRs) and undersampling ratios has been discussed as well.

It is very difficult to control a blast furnace efficiently because of the characteristics of the closed smelting and the system's large delay. It is very helpful for blast furnace control to establish a model that can accurately predict the tendency of silicon content in molten iron. In this article, a nonparallel hyperplane–based fuzzy classifier is proposed reconciling the model accuracy with the model interpretability. Each hyperplane is as close as possible to one class of samples and as far as possible from the other class of samples. This leads to two small quadratic programming problems instead of a large one as a support vector machine–based fuzzy classifier. This makes our classifier faster than a support vector machine–based fuzzy classifier. We evaluate the proposed method on blast furnace data and synthetic data, which verifies the effectiveness.

Abstract. This study tests the effects of camera distance and camera angle on emotional response across four categories of pictures covering a large emotional range (positive and negative miscellanea, erotica, and threat), using the International Affective Picture System (IAPS) –a large database of emotionally evocative photographs. We content analyzed 722 images for the content category and camera framing (distance and angle), employing these as independent factors in analyses, and used the IAPS’ pre-existing normative average ratings of emotional valence, arousal, and dominance as dependent variables. As hypothesized, affective responses were generally increased by closer framing and high and low angles (compared to straight angles), but the content of the picture played an important role in determining effect strength and direction. In particular, closeness increased arousal for all picture groups but had the opposite effect on positive miscellaneous pictures, straight angles decreased the emotional response for the two miscellanea groups, and low angles increased the emotional response for threatening pictures. This study is the first to show that previously found camera framing effects apply to pictures of high emotional intensity (e.g., erotica and threat). We suggest that future work should consider formal manipulations alongside message content.

Let us create a blank text file using the text editor or C++ IDE of our choice and name it source.cpp. First, let us create an empty C++ program that does nothing. The content of the source.cpp file is: int main(){}

The article proposes to consider the initial stage of the transition to such a form of training organization as collective classes in different-age groups of younger schoolchildren according to the methodology developed by the talented teacher A.G. Rivin. They are held at the Moscow secondary school named after Ivan Yarygin of the Krasnoyarsk region on the integrated content of academic subjects “Literary Reading”, “Fine Arts”, “Technology”. The specificity of the texts compiled by teachers for use in the process of paragraph-by-paragraph elaboration of the text using the technology of A.G. Rivina. Particular attention is paid to the issues of including first-graders (some of whom do not yet know how to read and write) in classes where younger schoolchildren of an older age participate. In this regard, multilevel algorithms of educational interaction are described. It highlights the difficulties that may arise with this form of training, as well as ways to overcome them.

Recently, the e-learners are drastically increased from the last two decades. Everything is learnt through internet without help of the tutor as well. For this purpose, the e-learners are required more e-learning applications that are able to supply optimal and satisfied data based on their capability. No content recommendation system is available for recommending suitable contents to the learners. For this purpose, this paper proposes a new semantic and fuzzy aware content recommendation system for retrieving the suitable content for the users. In this content recommendation system, we propose two content pre-processing algorithms namely Target Keyword based Data Pre-processing Algorithm (TKDPA) and Intelligent Anova-T Residual Algorithm (IAATRA) for selecting the more relevant features from the document. Moreover, a new Fuzzy rule based Similarity Matching algorithm (FRSMA) is proposed and used in this system for finding the similarity between the two terms and also rank them by using the newly proposed Similarity and Temporal aware Weighted Document Ranking Algorithm (STWDRA). In addition, a content clustering process is also incorporated for gathering relevant content. Finally, a new Fuzzy, Target Keyword and Similarity Score based Content Recommendation Algorithm (FTKSCRA) is also proposed for recommending the more relevant content to the learners accurately. The experiments have been conducted for evaluating the proposed content recommendation system and proved as better than the existing recommendation systems in terms of precision, recall, f-measure and prediction accuracy.

Besides NMR and X-ray crystallography, FTIR and CD spectroscopy are widely considered to be useful for determining protein secondary structure. These techniques can be used to obtain data in few minutes, using small quantities of proteins, which make them amenable for proteomics research. Here we explore the possibility of using artificial intelligence techniques to simultaneously analyse both FTIR and CD spectroscopic data for an identical set of proteins. Neural network analysis was carried out on normalised regions of FTIR (1700-1600 cm−1) and CD (180-259 nm) spectral data both with and without boxcar averaging in order to quantify the average length and percentages of secondary structures. A hybrid genetic algorithm/neural network approach, that automatically selects structure-sensitive wavelength/frequency, was used for the quantification of the protein secondary structure. Using this algorithm we also successfully identified the region of the CD spectrum that contains the most structure-sensitive information. This was located between 214-251 nm, suggesting that this region alone may be sufficient to rapidly determine the secondary structure content from CD spectral data. Overall, CD spectroscopic analysis produced better results compared to FTIR spectroscopy when selected wavelengths were used, although FTIR was better when the entire region between 1700-1600 cm−1 (FTIR), and 180-259 nm (CD), was subjected to neural network analysis. Application of Adaptive Neuro-Fuzzy Inference System (ANFIS) with fuzzy subtractive clustering for the analysis of the spectral data led to a slightly better prediction of the average helix/sheet length for FTIR spectroscopy compared to CD. Our findings reveal the potential of using artificial intelligence techniques for not only extracting structural information but also for better understanding of the relationship between complex spectral data and biologically important information.

The possibilities and feasibility of using the Bayesian network of trust and logistic regression to predict the content of nitrate nitrogen in the 0-40 cm soil layer before sowing have been investigated. Data from long-term multifactor field experience at the Siberian Research Institute of Farming and Agricultural Chemization of SFSCA RAS for 2013-2018 were used to train the models. The experiment was established on leached chernozem in the central forest-steppe subzone in 1981 in the Novosibirsk region. Considering the characteristics of the statistical sample (observation and analysis data), the main predictors of the models affecting nitrate nitrogen content in soil were identified. The Bayesian trust network is constructed as an acyclic graph, in which the main (basic) nodes and their relationships are denoted. Network nodes are represented by qualitative and quantitative plot parameters (soil subtype, forecrop, tillage, weather conditions) with corresponding gradations (events). The network assigns a posteriori probability of events for the target node (nitrate-nitrogen content in the 0-40 cm soil layer) as a result of experts completing the conditional probability table, taking into account the analysis of empirical data. Two scenarios were analyzed to test the sustainability of the network and satisfactory results were obtained. The result of the logistic regression is the coefficients characterizing the closeness of the relationship between the dependent variable and the predictors. The coefficient of determination of the logistic regression is 0.7. This indicates that the quality of the model can be considered acceptable for forecasting. A comparative assessment of the predictive capabilities of the trained models is given. The overall proportion of correct predictions for the Bayesian confidence network is 84%, for logistic regression it is 87%.

In recent years, high-dimensional histogram modification based reversible data hiding for images has drawn much attention among researchers. Pairwise prediction-error expansion (pairwise PEE) achieves great performance by exploiting the correlations among prediction-errors, in which a two-dimensional (2D) prediction-error histogram (PEH) is generated, and then modified based on a specific 2D modification mapping for reversible embedding. However, the used 2D mapping is fixed and heuristically designed without considering the image content, so the embedding performance of pairwise PEE can be further improved. To this end, the adaptive pairwise PEE (APPEE) is proposed in this paper to adaptively design the 2D mapping according to the distribution of 2D PEH, such that a better embedding performance is derived. To further improve the embedding performance, the proposed APPEE is extended from one single 2D PEH to multiple 2D PEHs generated based on different local complexities, in which multiple content-based 2D mappings are designed adaptively and individually for each 2D PEH. Experimental results show that the APPEE for either one single 2D PEH or multiple 2D PEHs significantly outperforms the conventional pairwise PEE. Simultaneously, the superiority of the proposed method is also experimentally verified compared with some other state-of-the-art works.

Nowadays, most research conducted in the field of abstractive text summarization focuses on neural-based models alone, without considering their combination with knowledge-based approaches that could further enhance their efficiency. In this direction, this work presents a novel framework that combines sequence-to-sequence neural-based text summarization along with structure and semantic-based methodologies. The proposed framework is capable of dealing with the problem of out-of-vocabulary or rare words, improving the performance of the deep learning models. The overall methodology is based on a well-defined theoretical model of knowledge-based content generalization and deep learning predictions for generating abstractive summaries. The framework is composed of three key elements: (i) a pre-processing task, (ii) a machine learning methodology, and (iii) a post-processing task. The pre-processing task is a knowledge-based approach, based on ontological knowledge resources, word sense disambiguation, and named entity recognition, along with content generalization, that transforms ordinary text into a generalized form. A deep learning model of attentive encoder-decoder architecture, which is expanded to enable a coping and coverage mechanism, as well as reinforcement learning and transformer-based architectures, is trained on a generalized version of text-summary pairs, learning to predict summaries in a generalized form. The post-processing task utilizes knowledge resources, word embeddings, word sense disambiguation, and heuristic algorithms based on text similarity methods in order to transform the generalized version of a predicted summary to a final, human-readable form. An extensive experimental procedure on three popular data sets evaluates key aspects of the proposed framework, while the obtained results exhibit promising performance, validating the robustness of the proposed approach.

MoS₂ negative-capacitance field-effect transistors (NCFETs) with a gate-stack of Hf₁₋ₓZrₓO₂/Al₂O₃ are fabricated, and the effects of the Zr content and the thickness of the stacked gate on electrical performance of the devices are investigated. It is found that a large improvement of device performance can be achieved by optimizing these parameters, e.g., using 6-nm Hf $_{0.5}$ Zr $_{0.5}$ O₂ (Zr content of 50%) as a ferroelectric layer and 2-nm Al₂O₃ as a matching layer, the relevant MoS₂ NCFET exhibits the excellent electrical characteristics: a low subthreshold swing (SS) of 31.4 mV/dec, a smaller counterclockwise hysteresis of ~75 mV, and a high on/off current ratio of 3.26 x 10⁶. The involved mechanisms lie in that: 1) an enhanced remnant polarization leads to low SS and 2) changing the thickness of the Hf₁₋ₓZrₓO₂ and Al₂O₃ can adjust the magnitude of ferroelectric capacitance and MOS capacitance to realize a reasonable match between both.

Spatial frequency domain imaging (SFDI) is a promising technique for its merits of noncontact and wide-field detection. However, considering the cost and detection speed, it has not been put into widespread application in agro-products. In this study, the low-cost, fast SFDI system and matching software were adopted to realize determination of absorption (μa) and reduced scattering coefficients (μ′s) of pears. The spatial frequencies (fx) were calibrated and system linearity was verified, the results showed that excellent linearity was obtained. Single snapshot demodulation was adopted, the validation results conducted on 390 liquid phantoms indicated that the optimal fx for snapshot method is 1/3 mm−1. Fast calculation models for μa and μ′s developed by least squares support vector regression (LSSVR) based on Monte Carlo (MC) simulations were then applied and validated at six wavelengths (460, 503, 527, 630, 658 and 675 nm). The results demonstrated that the LSSVR models could realize precise calculation for μa or μ′s. Finally, the variation trends of μa, μ′s, soluble solids content (SSC) and texture (MT firmness, flesh firmness, stiffness, brittleness and adhesiveness) of 9 batches of pears were analysed, and prediction models were developed by artificial neural network (ANN) based on μa and μ′s respectively. The results showed that for texture estimation, the prediction effect was relative well by using μ′s, especially for brittleness and adhesiveness, while the accuracy for SSC was limited by only six μa features. Future research should focus on the acquisition of more spectral information to improve model accuracy.