Specific Model Research Articles

Multi-Paragraph Machine Reading Comprehension with Hybrid Reader over Tables and Text

ABSTRACT In machine reading comprehension, answers to questions may be found in either text or tables. Previous studies have shown that table-specific pre-trained language models perform well when applied to tables; however, applying such models to input data that consists of both tables and text can be challenging. To address this issue, we introduce the hybrid reader model that can manage both tables and text using a modified K-Adapter architecture for effectively encoding the structured information of tables. The training process infuses knowledge for tabular data into a pre-trained model while retaining its original weights from pre-training. Hence, the pre-trained model is able to learn and utilize table information without sacrificing its previous training. Our proposed hybrid reader model achieved comparable or superior performance to that of a specialized model on the Korean MRC dataset, KorQuAD 2.0, using the provided adapters. Furthermore, we conducted experiments on an additional English MRC dataset and confirmed that our proposed model achieves performance comparable to that of the existing model. Our study indicates that employing a single hybrid model instead of two separate models can require fewer computing resources and less time while achieving comparable or superior performance, especially for techniques that apply projection and adapter.

Securing Breakneck Pace of 5G Networks Air Interfaces Through Proactive AI Monitoring

The promise of 5G networks enabling emerging technologies comes with formidable new security challenges. This paper proposes an AI-driven real-time security monitoring and incident response framework tailored to 5G infrastructure. A multi-layered architecture is presented using specialized deep learning models for radio access, edge, core, and network slice threat detection. Models including CNNs, RNNs, and transformers perform traffic analysis, signal classification, and log anomaly detection. A centralized controller aggregates model outputs into an integrated threat intelligence engine that deduces attack context and recommends mitigations. Further, a conversational bot interacts with security analysts in natural language to explain threats, suggest responses, and answer queries. The intelligent assistant is designed using dialog trees and transformer networks trained on security datasets. Evaluation on real-world 5G trial networks demonstrates 95% accuracy in classifying radio signal spoofing attacks and 98% precision in identifying malware infections. Analyst surveys confirm improved productivity and faster incident response with the AI assistant. As 5G matures, robust analytics and AI collaboration will grow increasingly critical for secure network operations. This research aims to provide both a conceptual framework and proven techniques as key enablers.

Learning Molecular Mixture Property Using Chemistry-Aware Graph Neural Network

Recent advances in machine learning (ML) are expediting materials discovery and design. One significant challenge facing ML for materials is the expansive combinatorial space of potential materials formed by diverse constituents and their flexible configurations. This complexity is particularly evident in molecular mixtures, a frequently explored space for materials, such as battery electrolytes. Owing to the complex structures of molecules and the sequence-independent nature of mixtures, conventional ML methods have difficulties in modeling such systems. Here, we present MolSets, a specialized ML model for molecular mixtures, to overcome the difficulties. Representing individual molecules as graphs and their mixture as a set, MolSets leverages a graph neural network and the deep sets architecture to extract information at the molecular level and aggregate it at the mixture level, thus addressing local complexity while retaining global flexibility. We demonstrate the efficacy of MolSets in predicting the conductivity of lithium battery electrolytes and highlight its benefits in the virtual screening of the combinatorial chemical space. Published by the American Physical Society 2024

Deep learning for quadratic hedging in incomplete jump market

We propose a deep learning approach to study the minimal variance pricing and hedging problem in an incomplete jump diffusion market. It is based on a rigorous stochastic calculus derivation of the optimal hedging portfolio, optimal option price, and the corresponding equivalent martingale measure through the means of the Stackelberg game approach. A deep learning algorithm based on the combination of the feed-forward and LSTM neural networks is tested on three different market models, two of which are incomplete. In contrast, the complete market Black–Scholes model serves as a benchmark for the algorithm’s performance. The results that indicate the algorithm’s good performance are presented and discussed. In particular, we apply our results to the special incomplete market model studied by Merton and give a detailed comparison between our results based on the minimal variance principle and the results obtained by Merton based on a different pricing principle. Using deep learning, we find that the minimal variance principle leads to typically higher option prices than those deduced from the Merton principle. On the other hand, the minimal variance principle leads to lower losses than the Merton principle.

Planar Lorentz invariant velocities with a wave equation application

In this paper we determine the functional form of those planar velocity fields for which the associated system of two ordinary differential equations are automatically invariant under a Lorentz transformation. For planar motion we determine first order partial differential equations for the velocity components u(x,y,t) and w(x,y,t) in the x− and y−directions respectively and their general solutions in terms of two arbitrary functions. These partial differential equations and the associated partial differential relations connecting energy and momentum are fully compatible with the Lorentz invariant energy–momentum relations and appear not to have been given previously in the literature. For a particular special relativistic model, one example is given involving similarity solutions of the wave equation. An interesting special case gives rise to a family of particle paths which are characterized by a single arbitrary function, and for which the magnitude of their velocities is the speed of light. This is indicative of the abundant possibilities existing in the “fast-lane”.

Fine Mapping of QTLs for Alkaline Tolerance in Crucian Carp (Carassius auratus) Using Genome-Wide SNP Markers.

Crucian carp (Carassius auratus) is widely distributed in the world and has become an economically freshwater fish. The population in Lake Dali Nur can tolerate the extreme alkaline environment with alkalinity over 50 mmol/L (pH 9.6), thus providing a special model for exploring alkali-tolerant molecular markers in an extremely alkaline environment. In this study, we constructed a high-density and high-resolution linkage map with 16,224 SNP markers based on genotyping-by-sequencing (GBS) consisting of 152 progenies and conducted QTL studies for alkali-tolerant traits. The total length of the linkage map was 3918.893 cM, with an average distance of 0.241 cM. Two QTLs for the ammonia-N-tolerant trait were detected on LG27 and LG45. A QTL for the urea-N-tolerant trait was detected on LG27. Interestingly, mapping the two QTLs on LG27 revealed that the mapped genes were both located in the intron of CDC42. GO functional annotation and KEGG enrichment analysis results indicated that the biological functions might be involved in the cell cycle, cellular senescence, MAPK, and Ras signaling pathways. These findings suggest that CDC42 may play an important role in the process of dealing with extremely alkaline environments.

Impulsive coupled systems with regular and singular ϕ-Laplacians and generalized jump conditions

This work contains sufficient conditions for the solvability of a third-order coupled system with two differential equations involving different Laplacians, fully discontinuous nonlinearities, two-point boundary conditions, and two sets of impulsive effects. The first existing result is obtained from Schauder’s fixed point theorem, and the second one provides also the localization of a solution via the lower and upper solutions technique.We point out that it is the first time that impulsive coupled systems with strongly nonlinear fully differential equations and generalized impulse effects are considered simultaneously. Moreover, the singular case is applied to a special relativity model in classical electrodynamics.

3D finite element modeling of fiber-reinforced pervious concrete with novel artificial intelligence supported image processing method

Pervious concrete is still being developed for stormwater management and flood risk reduction through experimental research and numerical modeling techniques. Due to its unique void structure, standard finite element solid modeling techniques need to be improved to represent this type of concrete, which requires special modeling techniques. This study presents a novel artificial intelligence-supported image processing method to model pervious concrete with varying fiber types and proportions realistically. The artificial intelligence-supported system initially reads the image sections obtained from the pervious concrete samples and processes these images. Afterward, many realistic three-dimensional (3D) finite element (FE) models were obtained, considering the void structure of the pervious concrete and the fiber types it contains with the developed coding. Laboratory tests were performed on specimens to evaluate the compressive and flexural strengths and to verify model adequacy. In addition, a constant-level permeability test was conducted to determine the permeability of the concrete. The accuracy of the models was confirmed by comparing the experimental results with 3D finite element analyses. The validation of this novel realistic modeling technique with laboratory experiments demonstrated that successful outcomes can be obtained using this method, particularly for other types of concrete with void or fiber-reinforced concretes that are complex to model.

On the sieve M-estimation for a special bilinear time series model with time-functional variance noises

. The bilinear model has been frequently used in fields like control theory and economics to model seismic data. In this article, time-functional variance (TFV) noises are embedded into a specific bilinear model. We propose a generalized autoregressive conditional heteroskedasticity-type maximum likelihood estimator (GMLE) with a sieve method and then provide various inferential techniques based on this GMLE. It is shown that under the finite fourth moment of errors, the GMLE is consistent and asymptotically normally distributed. A simulation study and analysis of real data are additionally carried out to evaluate GMLE’s finite sample performance.

Mayfly Algorithm for Modelling a Horizontal Flexible Plate Structure

Flexible plates are widely used in engineering and the industry, primarily due to the lightweight nature compared to rigid counterparts. These structures offer benefits such as cost savings, lower energy consumptions and improved operational safety. However, a notable drawback is that flexible structures are vulnerable to unwanted vibrations, which can cause structural damages. Hence, the development of specialized models are essential to effectively addressing this challenge. Researchers have devised various approaches to suppress unwanted vibrations, with contemporary studies often employing system identification techniques utilizing swarm intelligence algorithms to construct dynamic models of flexible structures. Therefore, this research employs the potent mayfly algorithm (MA), known for its effectiveness in optimization tasks. The developed models using MA were then compared with traditional approach known as recursive least square (RLS) through a comparative analysis. The outcome reveals that RLS exhibited the lowest mean square error (MSE) at , while MA had an MSE of Yet, MA adeptly depicted the characteristics of the system, outperforming the RLS in these validation by indicating a 95% confidence level in the correlation test and exhibiting robust stability in the pole-zero diagram. Consequently, MA serves as a fitting algorithm to accurately depict the real behaviour of the flexible plate structure.

P‐266: Late‐News Poster: Quality Assessment Towards Reflective Pattern Based on Diffraction Appearance

While polarizer‐free (Pol‐less) OLED displays increase the efficiency of the panels with lowering the thickness, there comes an issue of the screen quality drop due to diffraction. We have investigated phenomena and invented a novel model for diffraction, since the color‐difference‐based evaluation alone has limitations in responding to customer needs in various environments and reflecting situations in which actual users embrace the screen quality. The main spatial features of the diffraction appearance were classified into the size and complexity of the pattern, and each factor was quantified through a specialized model. Diffraction features can be automatically calculated as numerical values from the invented model reflecting perceptual levels, which acts as a ‘ruler’ to achieve a good 'grade' of image quality in the product development stage. In this paper, we demonstrate what plays vital roles in diffraction appearance on Pol‐less displays and how to evaluate the diffraction in representative values.

Statistical approaches enabling technology-specific assay interference prediction from large screening data sets

High throughput screening (HTS) technologies allow the biological testing of hundreds of thousands of compounds per day. Typically, a substantial proportion of the initial hits obtained by HTS are artifacts caused by assay interference. Therefore, global and technology-specific in silico models for identifying and predicting compounds interfering with biological assays have been developed. The global models benefit from training on large screening data sets, while the specialized models benefit from training on assay technology-specific experimental data. In this work, we develop and explore strategies for generating better predictors of technology-specific assay interference by utilizing the large bioactivity data matrices global models are trained on and employing partially new compound labeling approaches to maintain the assay technology awareness of specialized models. We demonstrate the utility of the statistically derived interference labels in machine learning using fluorescence-based assay interference as a representative example. Our random forest and multi-layer perceptron classifiers showed improved performance compared to existing models, achieving Matthews correlation coefficients (MCCs) of up to 0.47 on holdout data and up to 0.45 on an external test set. These results demonstrate that accurate assay-specific interference labels can be derived from large bioactivity data matrices, enabling the development of new machine-learning models without the need for further experimental data.

Zero-Shot Segmentation of Eye Features Using the Segment Anything Model (SAM)

The advent of foundation models signals a new era in artificial intelligence. The Segment Anything Model (SAM) is the first foundation model for image segmentation. In this study, we evaluate SAM's ability to segment features from eye images recorded in virtual reality setups. The increasing requirement for annotated eye-image datasets presents a significant opportunity for SAM to redefine the landscape of data annotation in gaze estimation. Our investigation centers on SAM's zero-shot learning abilities and the effectiveness of prompts like bounding boxes or point clicks. Our results are consistent with studies in other domains, demonstrating that SAM's segmentation effectiveness can be on-par with specialized models depending on the feature, with prompts improving its performance, evidenced by an IoU of 93.34% for pupil segmentation in one dataset. Foundation models like SAM could revolutionize gaze estimation by enabling quick and easy image segmentation, reducing reliance on specialized models and extensive manual annotation.

Interoperability Structure of Smart Water Conservancy Based on Internet of Things

Massive smart water conservancy object (WCO) need to be connected for real-time monitoring and control, which produces massive data. Unfortunately, heterogeneous data structures and semantics lead to low interoperability between WCO and management systems. To address this challenge, we propose a novel interoperability structure for a smart water conservancy system based on the Internet of Things (IoT), and the key design includes a smart WCO terminal, interoperability network, special interoperability protocol, WCO information model, and cloud platform. Universal terminal and network are the base of interoperability hardware, and special interoperability protocol and information model for interconnection of WCO are designed for smart water conservancy management system. WCO can be connected to a water conservancy Big Data processing cloud platform for interoperability applications. The application results demonstrate that our proposed WCO’s interoperability structure has obvious advantages than the general IoT at WCO interoperability. The interoperability protocol is reliable, the information model can ease interoperability and security, and the semantic dictionary is very rich and covers all semantic services of WCO.

CP factor model for dynamic tensors

Abstract Observations in various applications are frequently represented as a time series of multidimensional arrays, called tensor time series, preserving the inherent multidimensional structure. In this paper, we present a factor model approach, in a form similar to tensor CANDECOMP/PARAFAC (CP) decomposition, to the analysis of high-dimensional dynamic tensor time series. As the loading vectors are uniquely defined but not necessarily orthogonal, it is significantly different from the existing tensor factor models based on Tucker-type tensor decomposition. The model structure allows for a set of uncorrelated one-dimensional latent dynamic factor processes, making it much more convenient to study the underlying dynamics of the time series. A new high-order projection estimator is proposed for such a factor model, utilizing the special structure and the idea of the higher order orthogonal iteration procedures commonly used in Tucker-type tensor factor model and general tensor CP decomposition procedures. Theoretical investigation provides statistical error bounds for the proposed methods, which shows the significant advantage of utilizing the special model structure. Simulation study is conducted to further demonstrate the finite sample properties of the estimators. Real data application is used to illustrate the model and its interpretations.

Parametric Production of Prostheses Using the Additive Polymer Manufacturing Technology Multi Jet Fusion.

This study aims to develop a procedure for the production of 3D-printed forearm prostheses (especially hard outer sockets). The production procedure is designed in the form of a parametric workflow (CAD model), which significantly speeds up the designing process of the prosthesis. This procedure is not fixedly dependent on the software (SW) equipment and is fully transferable into another SW environment. The use of these prostheses will significantly increase the comfort of their patients' lives. It is possible to produce prostheses faster and in larger amounts and variants by the usage of additive technology. The input for the own production of the prosthesis is a model of the internal soft socket of the patient. This soft socket (soft bed) is made by a qualified prosthetist. A 3D-scanned CAD model is obtained afterward using the scanning method by an automatic laser projector. An editable, parametric external socket (modifiable in any CAD format) is generated from the obtained 3D scan using a special algorithmic model. This socket, after the necessary individual modifications, is transferred to 3D printing technology and produced using powder technology Multi Jet Fusion (HP MJF). The result of the designed and tested procedure is a quickly editable 3D-printed outer socket (main part of prosthesis), which is able to fully replace the current long-fiber composite solution. Production of current solutions is relatively time-consuming, and only one piece is produced in a given time. The newly designed technology eliminates this. This study summarized the possibilities of speeding up the production of forearm prostheses (but not only these) by creating a parametric CAD model that is applicable to different patients.

Legal security of mediation in Ukraine: shortcomings and prospects for improvement through the prism of the experience of the countries of the European Union

In the content of the article, the author investigated the peculiarities of the legal provision of mediation in Ukraine and the countries of the European Union. Despite the fact that today in legal science, a significant body of scientific works has been formed, which relate to certain aspects of mediation, it is worth noting that the issue of legal support for mediation in Ukraine has not been properly studied scientifically. Modern studies of the legal provision of mediation are especially relevant, given the high degree of its variability in Ukraine. On the basis of a critical analysis of the provisions of Ukrainian legislation in the field of mediation, its shortcomings and prospects for improvement were determined. The following are noted as shortcomings of the legal provision of mediation in Ukraine: a) the absence of provisions that would regulate the issues of mediation procedures, in particular before submitting an appeal to the court, during court proceedings, as well as during the execution of a court decision, etc.; b) inconsistency of its categorical and conceptual apparatus with established legislative terminology (for example, the terms "conflict” and "dispute”); c) inaccuracy of the legal provision of mediation principles; d) superficial, too general regulation of the mediation process itself; e) failure to take into account the European standard of mediation transparency, etc. The analysis of the peculiarities of the legal provision of the EU countries (France, Germany, Italy, Austria, etc.) made it possible to substantiate the prospects for improving the legal provision of mediation in Ukraine. It is summarized that the Ukrainian legislation has experienced legal transformations in the field of mediation, moving from the least effective and least reliable informal model to a specialized model by adopting the framework Law of Ukraine "On Mediation”. In our opinion, the most effective model of legal support for mediation is the specialized model chosen by Ukraine, which, among other things, should be complemented by effective mechanisms of self-regulation of this legal activity through the development and application of acts of the so-called "unofficial law” adopted by professional associations of mediators. The combination of the legal regulation of mediation in Ukraine was noted: a) the typicality of the law "On Mediation”, which established the legal foundation of mediation and regulated the most general and important issues in this area; b) self-regulating mechanisms of mediation, designed to ensure a quick response to the dynamic development of mediation and relations where it should be applied.

Treatment of phenolic wastewater in an anaerobic fluidized bed microbial fuel cell filled with graphene oxide-macroporous adsorption resin as multifunctional carrier

ABSTRACT A novel graphene oxide-modified resin (graphene oxide-macroporous adsorption resin) was prepared and used as a multifunctional carrier in an anaerobic fluidized bed microbial fuel cell (AFB-MFC) to treat phenolic wastewater (PW). The macroporous adsorption resin (MAR) was used as the carrier, graphene oxide was used as the modified material, the conductive modified resin was prepared by loading graphene oxide (GO) on the resin through chemical reduction. The modified resin particles were characterized by scanning electron microscopy (SEM), Raman spectroscopy (RS), specific surface area and pore structure analysis. Graphene oxide-macroporous adsorption resin special model was established using the Amorphous Cell module in Materials Studio (MS), and the formation mechanism of graphene oxide-macroporous adsorption resin was studied using mean square displacement (MSD) of the force module. Molecular dynamics simulation was used to study the motion law of molecular and atomic dynamics at the interface of graphene oxide-macroporous adsorption resin composites. The strong covalent bond between GO and MAR ensures the stability of GO/MAR. When the modified resin prepared in 3.0 mg/mL GO mixture was used in the AFB-MFC, the COD removal of wastewater was increased by 9.1% to 72.44%, the voltage was increased by 84.04% to 405.8 mV, and power density was increased by 765.44% to 242.67 mW/m2.

Wheat bread partially replaced with fermented cowpea flour: optimizing the formulation and storage study at 25 °C

Incorporating legume flours, like fermented cowpea flour (FCF), into wheat bread presents an opportunity to enhance protein intake. Fermenting cowpea flour through solid-state methods led to 79 % decrease in phytate levels, a 46 % reduction in tannins, and a 160 % increase in free amino acids. This study further aimed to optimize wheat bread by substituting wheat flour with FCF using a d-optimal mixture design. Varied proportions of wheat flour (66.3–93.7 %), FCF (5–30 %), xanthan gum (0.5–3 %), and lecithin (0.3–1 %) were tested to assess bread qualities such as hardness, color changes in crust and crumb, and specific loaf volume. A special cubic model was fitted to each response. The optimized blend comprised 75.2 % wheat flour, 22.9 % FCF, 1.6 % xanthan gum, and 0.3 % lecithin, scoring a desirability value of 0.63. The corresponding bread crumb hardness was 1981.3 N, and the loaf volume was 1.84 cm3/g. Shelf-life evaluation at 25 °C revealed microbial stability (aerobic mesophile count < 5 log10 cfu/g) of 6 days, with a decline in overall acceptability (9-point scale) ratings from 7.1 to 5.3 after this period. Over time, the bread's crumb experienced increased hardness and retrogradation enthalpy while moisture migration from crumb to crust was observed.

Incorporating Adaptive Sparse Graph Convolutional Neural Networks for Segmentation of Organs at Risk in Radiotherapy

Precisely segmenting the organs at risk (OARs) in computed tomography (CT) plays an important role in radiotherapy’s treatment planning, aiding in the protection of critical tissues during irradiation. Renowned deep convolutional neural networks (DCNNs) and prevailing transformer-based architectures are widely utilized to accomplish the segmentation task, showcasing advantages in capturing local and contextual characteristics. Graph convolutional networks (GCNs) are another specialized model designed for processing the nongrid dataset, e.g., citation relationship. The DCNNs and GCNs are considered as two distinct models applicable to the grid and nongrid datasets, respectively. Motivated by the recently developed dynamic-channel GCN (DCGCN) that attempts to leverage the graph structure to enhance the feature extracted by the DCNNs, this paper proposes a novel architecture termed adaptive sparse GCN (ASGCN) to mitigate the inherent limitations in DCGCN from the aspect of node’s representation and adjacency matrix’s construction. For the node’s representation, the global average pooling used in the DCGCN is replaced by the learning mechanism to accommodate the segmentation task. For the adjacency matrix, an adaptive regularization strategy is leveraged to penalize the coefficient in the adjacency matrix, resulting in a sparse one that can better exploit the relationships between nodes. Rigorous experiments on multiple OARs’ segmentation tasks of the head and neck demonstrate that the proposed ASGCN can effectively improve the segmentation accuracy. Comparison between the proposed method and other prevalent architectures further confirms the superiority of the ASGCN.