Search Results (277)

The evolution of future Internet and sixth-generation (6G) networks is driving a paradigm shift from classical bit-centric communication toward meaning-aware and task-oriented communication models. Traditional information theory, while fundamental for ensuring reliable symbol transmission, does not account for semantic relevance or task effectiveness, which are critical for emerging applications such as autonomous systems, immersive services, and ultra-low-latency communications. This article presents a comprehensive review of Semantic Communications (SemCom) from a future Internet perspective. The review systematically analyses representative extensions of classical information theory aimed at quantifying semantic information, including semantic information measures, semantic channel capacity, and semantic rate–distortion formulations. In addition, the main mathematical and computational frameworks enabling practical semantic communication systems are examined, including the Information Bottleneck principle, learning-based end-to-end communication architectures, and reinforcement learning approaches for task-oriented optimization under network constraints. The review further discusses the role of semantic metrics, contextual modelling, and task-driven performance evaluation in the design of semantic-aware communication systems. The analysis identifies key open challenges, particularly the lack of a unified theoretical framework, the need for robust and context-aware semantic performance metrics, and the integration of semantic awareness into network-level design. Overall, this review highlights Semantic Communications as a promising paradigm for future Internet and 6G networks, where communication efficiency is increasingly determined by semantic relevance and task effectiveness rather than bit-level fidelity alone. Full article

This study stems from the need to investigate how GenAI tools, particularly ChatGPT-4o, can support the professional development of mathematics teachers. It explores how Blum’s modeling cycle can serve as a conceptual and operational framework for mathematics teachers’ instructional design when supported by ChatGPT-4o. Drawing on a qualitative case study within a teacher professional development program, the research analyzes how two upper secondary school teachers engaged with ChatGPT-4o to redesign a mathematical task involving probability and real-world contexts. Data include responses to three modeling-related tasks, teachers’ prompts and interactions with ChatGPT-4o, and the final mathematical activity they designed. These materials were analyzed qualitatively according to the modeling cycle and its sub-competencies. The results indicate that the modeling cycle provided teachers with a cognitive and methodological scaffold to guide their interaction with ChatGPT-4o, allowing them to structure, validate, and refine AI-generated ideas through all stages of modeling—from understanding and mathematizing to interpreting and validating. These findings suggest that the modeling cycle can be reinterpreted as a design-oriented framework for integrating ChatGPT-4o in mathematics teacher education. Implications for teacher professional development and future research directions are discussed. Full article

Ensuring the required microclimate parameters is the most critical task in hot climates. In pig farms, air cooling is provided by means of steam-compression chillers or evaporative cooling, which is the simplest way to cool the air. The implementation of evaporative cooling depends largely on the interaction of the media involved in this process. This paper considers the process of interaction of cooling water with the surface of a cellular polycarbonate heat exchanger. A mathematical model describing the process of wetting the sprayed surface of the heat exchanger is obtained. The authors determined the theoretical water flow rate required to provide air cooling for a given operation mode. Experimental trials of a recuperative heat recovery unit with a heat exchanger made of cellular polycarbonate equipped with a water evaporative cooling system were carried out. The authors conducted a comparative assessment to evaluate the effectiveness of evaporative cooling in a heat recovery unit equipped with a polycarbonate heat exchanger versus panel evaporative systems using wetted paper pads at pig farms in the Vladimir and Tambov regions of Russia. The panel evaporative coolers provided a temperature reduction of 11.3 °C without any splashing effect. Under the same operating conditions, the heat recovery unit achieved an inlet air temperature reduction of 10.5 °C, accompanied by splashing. When the water flow rate supplied for evaporation was reduced until the splashing ceased, the cooling temperature drop decreased to 10.1 °C, which is 11% lower, compared with the paper pads. The study revealed characteristic operating modes for the unit that ensure effective air cooling, depending on the cooling water flow rate. Since the prevailing temperature during the system’s main operating time is significantly lower than the design temperature (the absolute temperature maximum), to achieve effective cooling of the supply air without splashing or excessive water waste, the cooling circuit water should circulate at a flow rate within 40 to 63% of the maximum design value. Alternatively, an automated control system should be employed to regulate the water supply based on outdoor air temperature and humidity. Full article

Deep learning approaches for egocentric video understanding often lack a principled theoretical treatment of stability, particularly when dealing with the sparse, noisy, and temporally ambiguous observations characteristic of first-person imaging. In this work, we frame egocentric video question answering not merely as a classification task, but as an ill-posed inverse problem aimed at reconstructing latent semantic intent from stochastically perturbed visual signals. To address the instability inherent in standard dual-encoder architectures, we present a framework with a mathematical interpretation that incorporates gated cross-modal interaction within the transformer backbone. Formally, the video-side update analyzed in this work is defined as a learnable convex combination of unimodal feature representations and cross-modal attention residuals; the full implementation applies analogous gated cross-modal updates bidirectionally. From a regularization perspective, the gating mechanism can be interpreted as an adaptive parameter that balances data fidelity against language-conditioned structural constraints during feature reconstruction. We provide the Bounded Update Property (Lemma 1) and an analytical layer-wise sensitivity bound and empirically demonstrate that the proposed framework achieves measurable improvements in both accuracy and stability on the EgoTaskQA and MSR-VTT benchmarks. On EgoTaskQA, our model improves accuracy from 27.0% to 31.7% (+4.7 pp) and reduces the accuracy drop under 50% frame drop from 3.93 pp to 0.94 pp. On MSR-VTT, our model improves accuracy by 13.0 pp over the dual-encoder baseline. Under severe perturbation (50% frame drop) on MSR-VTT, our model retains 97.7% of its clean performance, whereas the baseline exhibits near-zero drop accompanied by majority-class behavior. These results provide empirical evidence that the proposed interaction induces stable behavior under perturbations in an ill-posed multimodal inference setting, mitigating sensitivity to sampling variability while preserving query-relevant temporal structure. Furthermore, an entropy-based analysis indicates that the gating mechanism prevents excessive diffusion of attention, promoting coherent temporal reasoning. Overall, this work offers a mathematically informed perspective on designing interaction mechanisms for stable multimodal systems, with a focus on robust reasoning under temporal ambiguity. Full article

Most pre-trained language representation models are designed to encode contextualized semantic information for general language processing tasks. However, they are insufficient for math word problem (MWP) solving, which requires not only linguistic syntax and semantic understanding but also numerical reasoning. In this work, we introduce SSN4Solver, a deep neural solver that improves MWP-solving performance by symmetrically fusing syntax, semantics, and numeracy representations within its contextual encoder. Our approach jointly captures syntactic structures from dependency trees, semantic features from part-of-speech tags, and the attributes and relations of numerical entities. By treating these heterogeneous information sources in a balanced and aligned manner, SSN4Solver constructs a rich, multi-faceted representation for MWP solving without introducing substantial computational overhead, empowering human–computer interaction (HCI) applications such as adaptive educational interfaces and intelligent tutoring systems. Extensive experiments demonstrate that SSN4Solver outperforms existing baseline models. In addition, a visualization scheme is designed to elucidate how the three types of representations contribute to the solving process. SSN4Solver thus offers a scalable solution, contributing to the development of HCI systems that are both intelligent and mathematically effective. Full article

This study evaluates whether theory-informed, mathematically focused support can ease the school-to-university transition in an unequal South African STEM context. First-year students could voluntarily attend Mathematical Thinking Workshops (MTWs) grounded in constructivism, the zone of proximal development, APOS theory, and cognitive load theory, providing low-threat, collaborative practice with non-routine, representation-rich tasks. Because attendance was self-selected, we used a quasi-experimental design: participation was modeled from pre-university covariates (school-leaving Mathematics and English grades and standardized university preparedness tests in Mathematics and Quantitative Literacy), and MTW participants were matched to comparable non-participants using nearest-neighbor propensity-score matching. Average treatment effects on the treated were estimated for multiple assessments and for a composite score capturing performance on higher-order items within those assessments. MTW participants outperformed matched peers on early first-semester assessments, especially those containing the most higher-order items, indicating that workshops helped when cognitively demanding tasks first appeared. Effects on later, more distal assessments were positive but attenuated, producing an “early gains, fading effects” pattern. Although estimates were imprecise, benefits appeared largest for students who had scored 70–84% in school-leaving mathematics. Overall, the findings suggest that transitional workshops can deliver timely, assessment-visible gains, although these effects may weaken over time when they are not reinforced or well aligned with later summative assessment. Full article

Resource-constrained intelligent systems increasingly require reliable long-form reasoning capabilities under limited computational and memory budgets, particularly in edge and embedded sensing environments. However, reinforcement learning for long-horizon decision generation remains highly unstable in such low-resource settings due to severe reward sparsity and imbalanced credit assignment, which often lead to non-convergent or excessively verbose generation behavior. In this work, we propose AF-CuRL (Answer-Focused Curriculum Reinforcement Learning), a lightweight reinforcement learning framework designed to stabilize long-form generation without increasing model size or computational cost. AF-CuRL improves optimization learnability through two complementary objective-level designs: (1) answer-focused token reweighting, which concentrates policy updates on reward-critical regions of generated sequences to alleviate credit assignment imbalance, and (2) a two-phase curriculum reward schedule that prioritizes stable termination and output regularity before shifting toward correctness-oriented optimization. We evaluate AF-CuRL on a 1.5B-parameter language model under strictly constrained training settings, using mathematical reasoning tasks as a controlled and reproducible proxy for long-horizon, rule-based decision-making commonly encountered in intelligent sensing and embedded systems. Experimental results demonstrate consistent improvements in both decision accuracy and generation regularity, including higher termination reliability and reduced generation length, compared with standard sequence-level reinforcement learning baselines. These results suggest that, for resource-limited and edge-intelligent systems, structured objective design can be more effective than model scaling for achieving stable and efficient long-form reasoning, providing a practical reinforcement learning solution for intelligent systems operating under real-world constraints. Full article

Industrial robots are a widely used technology in various industries, capable of performing multiple tasks, particularly machining operations. However, compared to CNC machines, industrial robots have certain drawbacks, such as lower rigidity, precision, and repeatability. These limitations are significant obstacles to fully integrating industrial robots into manufacturing processes. The deficiencies in robots directly impact the quality of machined workpieces, leading to reduced surface quality and geometric accuracy. Researchers have been exploring ways to improve production accuracy and quality using robotic arms for decades. Despite technological advancements, the full implementation of robotic arms in machining processes remains in its early stages. This article focuses on robotic milling and the effect of technological parameters on the final quality of the machined material, aiming to determine their importance for designing the main experimental plan. In our experiment, we used an ABB IRB 1100 robotic arm with a payload of 4 kg and a maximum reach of 0.58 m. The milling process was programmed in RobotStudio, which also supports milling complex shapes. Our statistical analysis showed that feed rate had the greatest impact on quality, while depth of field had the least influence. This information will guide further research in developing a mathematical model for robotic milling. Full article

This study explored how mathematics teachers in Chinese special schools provide instructional support to primary-aged students with intellectual and developmental disabilities (IDD). The types, characteristics, and classroom implementation processes of such support were identified to address a gap in the literature regarding subject-specific instructional practices in special education settings. A qualitative research design using interpretative phenomenological analysis (IPA) was employed. Five mathematics teachers from special schools in Shanghai participated in the study. Data were collected through 15 video-recorded classroom observations and five semi-structured interviews. Thematic analysis was conducted to identify key patterns of instructional support. The analysis revealed five core domains of instructional support for students with IDD: (1) comprehension facilitation through simplified explanations, real-life connections, and visual scaffolding; (2) responding to tasks involving prompts, modeling, and hand-over-hand support; (3) maintaining attention using individual and collective cues; (4) sustaining motivation through praise, encouragement, and second-chance opportunities; and (5) regulating behavior such as verbal restraint, physical proximity, and attention redirection. The findings contribute to a deeper understanding of effective instructional support tailored to students with IDD. Full article

Recent years have witnessed performance improvements in Multimodal Large Language Models (MLLMs) on downstream natural image understanding tasks. However, when applied to the function graph reasoning task, which is highly information-dense and abundant in fine-grained structural details, these models face pronounced performance degradation. The challenges are primarily characterized by several core issues: the static projection bottleneck, inadequate cross-modal interaction, and insufficient visual context in text embeddings. To address these problems, this study proposes a multimodal feature fusion enhancement method for function graph reasoning and constructs the FuncFusion-Math model. The core innovation of this model resides in its design of a dual-path feature fusion mechanism for both image and text. Specifically, the image fusion module adopts cross-attention and self-attention mechanisms to optimize visual feature representations under the guidance of textual semantics, effectively mitigating fine-grained information loss. The text fusion module, through feature concatenation and Transformer encoding layers, deeply integrates structured mathematical information from the image into the textual embedding space, significantly reducing semantic deviation. Furthermore, this study utilizes a four-stage progressive training strategy and incorporates the LoRA technique for parameter-efficient optimization. Experimental results demonstrate that the FuncFusion-Math model, with 3B parameters, achieves an accuracy of 43.58% on the FunctionQA subset of the MathVista test set, outperforming a 7B-scale baseline model by 13.15%, which validates the feasibility and effectiveness of the proposed method. Full article

Reliable posture and loading characteristics detection of hydraulic supports is one of the indispensable factors to realizing the intelligentization of fully mechanized coal mining faces. Due to the complexity and dynamic nature of mining process, achieving real-time and accurate detection of the hydraulic support posture and load presents an exceptionally challenging task. Therefore, an interactive algorithm for evaluating the load-carrying characteristic of hydraulic support by considering the three-dimensional space driving theory and dynamic theory was developed and experimentally verified based on a self-designed experimental platform. The paper aimed to establish a three-dimensional spatial dynamic and kinematics model for shield support, evaluating its loading performance in challenging working conditions. Initially, a three-dimensional kinematics model was developed to describe the bearing capacity of powered support in various postures based on the three-dimensional drive space theory. A dynamic model was suggested to investigate the effects of multiple factors on the position of hydraulic support drive units on their load-carrying capability in various demanding working situations. The results indicate that increasing the length of the drive units can significantly improve the bearing performance of shield support. The proposed mathematical technique offers a novel method for modifying the coupling of surrounding rock with hydraulic supports and supplying coal mining with real-time assistance. Full article

To enhance desalination efficiency and reduce experimental costs, the development of advanced mathematical models for EMS is essential. In this study, we propose a novel hybrid approach that integrates neural networks with high-accuracy numerical simulations of electroconvection. Based on dimensionless similarity criteria (Reynolds, Péclet numbers, etc.), we establish functional relationships between critical parameters, such as the dimensionless electroconvective vortex diameter and the plateau length of current–voltage curves. Training datasets were generated through extensive numerical experiments using our in-house developed mathematical model, while multilayer feedforward neural networks with backpropagation optimization were employed for regression tasks. The resulting AI (artificial intelligence)-driven hybrid models enable rapid prediction and optimization of EMS design and operating parameters, reducing computational and experimental costs. This research is situated at the intersection of membrane science, artificial intelligence, and computational modeling, forming part of a broader foresight agenda aimed at developing next-generation intelligent membranes and adaptive control strategies for sustainable water treatment. The methodology provides a scalable framework for integrating physically based modeling and machine learning into the design of high-performance electromembrane systems. Full article

Existing experience-based methods cannot effectively assist commercial building operators in allocating workforce resources according to contracts and balance multiple workforce management objectives under resource constraints, leading to misaligned allocation strategies. To address this issue, this study develops a workforce resource allocation optimization model based on BERT and the NSGA-II. First, a natural language processing (NLP) model is trained to extract operational tasks from contracts and match required workforce types, thereby establishing the framework for workforce allocation schemes. Second, a mathematical optimization model for workforce allocation strategies is constructed with the objectives of minimizing workforce wage costs (B1), maximizing average service levels (B2), and maximizing average digital technology acceptance (B3). An algorithm based on NSGA-II is then designed to solve the model and obtain the optimal Pareto solution set of allocation schemes. Third, the CRITIC–VIKOR method evaluates the Pareto set and determines the final recommended schemes. A case study was conducted on a university campus in Shandong, China, to validate the model’s effectiveness. The results show that the NLP model successfully identified 14 operational tasks and 13 required workforce types from the contract. Compared with the operator’s expected values (B1 = 46,0000 CNY, B2 = 65 points, B3 = 50 points), the optimal allocation scheme calculated using NSGA-II and the CRITIC–VIKOR method reduces B1 by 10.79%, increases B2 by 18.02%, and improves the B3 by 16.79%. This study formulates the workforce allocation problem in the operation stage as a mathematical optimization model and, for the first time, incorporates the workforce’s digital technology acceptance as an optimization objective, thereby filling a theoretical gap in workforce management for commercial building operations. The proposed model provides operators with a semi-automated decision-support tool to enhance workforce management, thereby promoting the sustainable operation of commercial buildings. Full article

Assessment practices in Romanian primary education remain largely recall-based, despite curriculum expectations that prioritize reasoning, metacognition, and inclusive learning processes. This conceptual–analytical study examines the structural misalignments between curriculum goals, classroom assessment cultures, and national evaluation systems, highlighting their impact on learning equity and cognitive development. Drawing on international frameworks (OECD, UNESCO), national assessment data, and Romanian pedagogical literature, the analysis identifies three systemic gaps: curriculum–assessment misalignment, assessment–instruction misalignment, and a mismatch between equity-oriented policies and classroom practice. To address these challenges, the article proposes the ECEI Framework, an integrated developmental model that combines principles of cognitive education, metacognitive strategy development, inclusive pedagogy, and formative assessment. The framework introduces four categories of indicators—cognitive, metacognitive, inclusive, and assessment—designed to support teachers in observing and evaluating learning processes more effectively in diverse classrooms. Discipline-based illustrations in mathematics, reading, and science demonstrate how innovative assessment practices can make students’ thinking visible through authentic tasks, learning progressions, and multimodal response pathways. The findings suggest that developmental and inclusive assessment is essential for improving learning outcomes and reducing socio-economic disparities in primary education. Implementing the ECEI Framework requires targeted teacher training, coherent curriculum–assessment alignment, and system-level support to ensure sustainable changes in instructional practice. Full article

The development of neuroanatomy and neurophysiology has revealed many new details about neurons’ operation over the past few decades, requiring modifications to their theoretical models. The development of computing technology enables us to consider the fine details the new model requires, but it necessitates a different approach. To achieve that goal, the disciplinarity of science must be revisited for living matter, the theoretical model must be updated, and a series of processes instead of states must be considered; furthermore, new mathematics, algorithms, and computing technologies for the new view are also needed. We provide an algorithm implementing the mathematics of the updated theoretical model that considers the neuronal current to consist of charged ions (and so considers thermodynamic effects) and opens the way for explaining the mechanical, optical, etc., consequence phenomena of the electrical operation. We use a new technology in this effort: a tool designed to achieve extreme accuracy in simulating high-speed electronic circuits. The algorithm applies the cross-disciplinary unified electrical/thermodynamic model, along with an unusual programming method, to provide new insights into neuronal operations, describe the processes that take place in living matter, and determine their computing implementation. As has long been suspected, the faithful simulation of biological processes requires accurately mapping biological time to technical computing time. Therefore, the paper focuses on time handling in biology-targeting computations, especially in large-scale tasks. We also touch on the question of simulating the operation of their network, which is contrasted with that of Spiking Neural Networks. The way technical computing works inhibits efforts to achieve the required accuracy in reproducing the temporal behavior of biological operations using conventional computer programs. Full article