Representation Method Research Articles

Electronic properties and vibrationally averaged structures of X̃ 2Σ+ MgOH: a computational molecular spectroscopy study.

For X̃ 2Σ+ MgOH, we have calculated the 3D potential energy surface (PES) at the MR-SDCI+Q/[cc-pCV5Z (Mg), aug-cc-pV5Z (O), cc-pV5Z (H)] level and derived the vibrational properties from there using the discrete variable representation (DVR) method. The PES minimum is at the linear structure; hence, MgOH is a "linear molecule." The 3D PES is shallow, and MgOH tends to bend in the region immediately when either or both Mg-O and O-H bonds become longer than those of the equilibrium structure (re(Mg-O) = 1.7614 Å, re(O-H) = 0.9453 Å, and ∠e(Mg-O-H) = 180°). The zero-point structure, determined as the expectation values over the DVR3D wavefunctions, has 〈r(Mg-O)〉0 = 1.7837 Å and 〈r(O-H)〉0 = 0.9948 Å, and the deviation angle from linearity 〈〉0 = 26.4°. The harmonic frequencies ωe for the Mg-O stretching, bending, and O-H stretching modes are 768, 142, and 4060 cm-1, respectively, and the corresponding term values ν1, ν2, and ν3 are 752, 156, and 3867 cm-1. All the vibrational behaviors, such as quasi-linear features, unusual relationship ν2 > ω2, a large amplitude bending motion, etc., are elucidated in terms of the ab initio electronic wavefunctions and the DVR3D vibrational wavefunctions. We have another piece of evidence to support our postulation that a linear molecule is observed as being bent.

Overview of knowledge graph construction in the field of fully electronic computer interlocking systems

As a core technology in railway signal control, the fully electronic computer interlocking system achieves the control of railway signals and switches through computerization, ensuring the safety and efficiency of train operations. With the increasing complexity of railway systems, traditional management and maintenance face significant challenges. Knowledge graphs, as an advanced data management and representation method, can effectively organize and utilize large-scale domain knowledge, enhancing the system's intelligence. This paper introduces the construction methods, core concepts, and applications of knowledge graphs in the field of fully electronic computer interlocking systems, providing references for research and practice in this area.

Three minimal norm Hermitian solutions of the reduced biquaternion matrix equation EM+M˜F=G$$ EM+\tilde{M}F=G $$

In this paper, we investigate the minimal norm Hermitian solution, pure imaginary Hermitian solution and pure real Hermitian solution of the reduced biquaternion matrix equation. We introduce the new real representation of the reduced biquaternion matrix and the special properties of . We present the sufficient and necessary conditions of three solutions and the corresponding numerical algorithms for solving the three solutions. Finally, we show that our method is better than the complex representation method in terms of error and CPU time in numerical examples.

Identification of Optimal Machine Learning Algorithms and Molecular Fingerprints for Explainable Toxicity Prediction Models Using ToxCast/Tox21 Bioassay Data.

Recent studies have primarily focused on introducing novel frameworks to enhance the predictive power of toxicity prediction models by refining molecular representation methods and algorithms. However, these methods are inherently complex and often pose challenges in understanding and explaining, leading to barriers in their regulatory adoption and validation. Therefore, it is necessary to select the optimal model, considering not only model performance but also interpretability. This study aimed to identify the optimal combination of molecular fingerprints (pattern-based versus algorithm-based) and machine learning algorithms (simple versus complex) for developing explainable toxicity prediction models through an comprehensive investigation of the ToxCast/Tox21 bioassay data set. For 1092 ToxCast/Tox21 assays, five molecular fingerprints (MACCS, Morgan, RDKit, Layered, and Patterned) and six algorithms (MLP, GBT, Random Forest, kNN, Logistic Regression, and Naïve Bayes) were used to train the models. Results showed that 35 models revealed acceptable performance (F1 score or accuracy is 0.8 or higher). Among the combinations, either MACCS or Morgan, paired with Random Forest, demonstrated robust performance compared with other molecular fingerprints and algorithms. MACCS and Random Forest are valuable, even when prioritizing interpretability. Consequently, the MACCS-Random Forest combination model based on four assays, targeting G protein-coupled receptor and kinase, were identified and they can be used to discern specific structural features or patterns in chemical compounds, offering explainable insights into toxicity-related chemical structures. This study indicates the importance of not disregarding the utilization of simple models when assessing both predictivity and interpretability within the context of chemical feature-based Tox21 data analysis.

ChronoVectors: Mapping Moments through Enhanced Temporal Representation

Time-series data are prevalent across various fields and present unique challenges for deep learning models due to irregular time intervals and missing records, which hinder the ability to capture temporal information effectively. This study proposes ChronoVectors, a novel temporal representation method that addresses these challenges by enabling a more specialized encoding of temporal relationships through the use of learnable parameters tailored to the dataset’s dynamics while maintaining consistent time intervals post-scaling. The theoretical demonstration shows that ChronoVectors allow the transformed encoding tensors to map moments in time to continuous spaces, accommodating potentially infinite extensions of the sequence and preserving temporal consistency. Experimental validation using the Parking Birmingham and Metro Interstate Traffic Volume datasets reveals that ChronoVectors enhanced the predictive capabilities of deep learning models by reducing prediction error for regression tasks compared to conventional time representations, such as vanilla timestamp encoding and Time2Vec. These findings underscore the potential of ChronoVectors in handling irregular time-series data and showcase its ability to improve deep learning model performance in understanding temporal dynamics.

Leveraging multiple behaviors and explicit preferences for job recommendation

Numerous job openings have been posted online, indicating the increasing importance of job recommendations. Recently, job seekers often enter their preferences into job search websites to receive some job recommendations that they hope to apply for. To achieve this goal, the following two types of data are available: (1) auxiliary behavior data such as viewing job postings, bookmarking them and (2) explicit preference data such as conditions for a job that each job seeker desires. Although limited research has employed both (1) and (2) simultaneously, no sophisticated job recommendation method leverages multiple types of interactions and explicit preferences to achieve high accuracy. Given this point, we propose a method for job recommendation that employs auxiliary behavior data and each user’s explicit preference data simultaneously. Additionally, our proposed method addresses multiple behavior overlaps and refines the latent representations. Furthermore, the integration method of the latent representations obtained from each of the two modules addresses the consistency of user preferences and the similarity with job postings, enabling a more accurate estimation of user preferences. Experimental results on our dataset constructed from an actual job search website show that our proposed model outperforms the best baseline by 31.4% and 32.8% in terms of Mean Reciprocal Rank (MRR) and Normalized Discounted Cumulative Gain@5 (nDCG@5), respectively. We have released our source codes.11https://github.com/saitoxu/JME-ESWA.

Data-Driven Strain Sensor Design Based on a Knowledge Graph Framework.

Wearable flexible strain sensors require different performance depending on the application scenario. However, developing strain sensors based solely on experiments is time-consuming and often produces suboptimal results. This study utilized sensor knowledge to reduce knowledge redundancy and explore designs. A framework combining knowledge graphs and graph representational learning methods was proposed to identify targeted performance, decipher hidden information, and discover new designs. Unlike process-parameter-based machine learning methods, it used the relationship as semantic features to improve prediction precision (up to 0.81). Based on the proposed framework, a strain sensor was designed and tested, demonstrating a wide strain range (300%) and closely matching predicted performance. This predicted sensor performance outperforms similar materials. Overall, the present work is favorable to design constraints and paves the way for the long-awaited implementation of text-mining-based knowledge management for sensor systems, which will facilitate the intelligent sensor design process.

Non-Intrusive System for Honeybee Recognition Based on Audio Signals and Maximum Likelihood Classification by Autoencoder.

Artificial intelligence and Internet of Things are playing an increasingly important role in monitoring beehives. In this paper, we propose a method for automatic recognition of honeybee type by analyzing the sound generated by worker bees and drone bees during their flight close to an entrance to a beehive. We conducted a wide comparative study to determine the most effective preprocessing of audio signals for the detection problem. We compared the results for several different methods for signal representation in the frequency domain, including mel-frequency cepstral coefficients (MFCCs), gammatone cepstral coefficients (GTCCs), the multiple signal classification method (MUSIC) and parametric estimation of power spectral density (PSD) by the Burg algorithm. The coefficients serve as inputs for an autoencoder neural network to discriminate drone bees from worker bees. The classification is based on the reconstruction error of the signal representations produced by the autoencoder. We propose a novel approach to class separation by the autoencoder neural network with various thresholds between decision areas, including the maximum likelihood threshold for the reconstruction error. By classifying real-life signals, we demonstrated that it is possible to differentiate drone bees and worker bees based solely on audio signals. The attained level of detection accuracy enables the creation of an efficient automatic system for beekeepers.

Enhanced multi-scenario running safety assessment of railway bridges based on graph neural networks with self-evolutionary capability

Accurate and efficient safety assessment for train-bridge coupled (TBC) systems is paramount in railway engineering. Traditional neural networks, though efficient and apt for real-time applications, falter in generalizing to untrained engineering scenarios. Addressing this gap, this study introduces a novel graph isomorphic network (GIN) endowed with the self-evolutionary capability, specifically tailored for seismic safety assessments across various scenarios in TBC systems. Initially, the study devises a graph representation method that adeptly encapsulates the structural complexities of TBC systems into graph data. This data is subsequently processed by the enhanced GIN model, refined for optimal alignment with engineering structures. The unique self-evolutionary mechanism of the GIN model equips it to interpret and predict the characteristics of new engineering scenarios, leveraging the extant training data. Empirical evaluations reveal the proposed GIN model's superiority in assessing known scenarios and its exceptional ability to generalize to new ones, outperforming traditional neural networks. Additionally, this study underscores the GIN model's potential applicability across diverse engineering fields.

Random memristor-based dynamic graph CNN for efficient point cloud learning at the edge

The broad integration of 3D sensors into devices like smartphones and AR/VR headsets has led to a surge in 3D data, with point clouds becoming a mainstream representation method. Efficient real-time learning of point cloud data on edge devices is crucial for applications such as autonomous vehicles and embodied AI. Traditional machine learning models on digital processors face limitations, with software challenges like high training complexity, and hardware challenges such as large time and energy overheads due to von Neumann bottleneck. To address this, we propose a software-hardware co-designed random memristor-based dynamic graph CNN (RDGCNN). Software-wise, we transform point cloud into graph, and propose random EdgeConv for efficient hierarchical and topological features extraction. Hardware-wise, leveraging memristor’s intrinsic stochasticity and in-memory computing capability, we achieve significant reductions in training complexity and energy consumption. RDGCNN demonstrates high accuracy and efficiency across various point cloud tasks, paving the way for future edge 3D vision.

Expansion by regions meets angular integrals

We study the small-mass asymptotic behavior of so-called angular integrals, appearing in phase-space calculations in perturbative quantum field theory. For this purpose we utilize the strategy of expansion by regions, which is a universal method both for multiloop Feynman integrals and various parametric integrals. To apply the technique to angular integrals, we convert them into suitable parametric integral representations, which are accessible to existing automation tools. We use the code asy.m to reveal regions contributing to the asymptotic expansion of angular integrals. To evaluate the contributions of these regions in an epsilon expansion we apply the method of Mellin-Barnes representation. Our approach is checked against existing results on angular integrals revealing a connection between contributing regions and angular integrals constructed from an algebraic decomposition. We explicitly calculate the previously unknown asymptotics for angular integrals with three and four denominators and formulate a conjecture for the leading asymptotics and the pole part for a general number of denominators and masses.

Mileage-Aware for Vehicle Maintenance Demand Prediction

It is of paramount importance to accurately predict the maintenance demands of vehicles in order to guarantee their sustainable use. Nevertheless, the current methodologies merely predict a partial aspect of a vehicle’s maintenance demands, rather than the comprehensive maintenance demands. Moreover, the process of predicting vehicle maintenance demands must give due consideration to the influence of mileage on such demands. In light of the aforementioned considerations, we put forth a vehicle overall maintenance demand prediction method that incorporates vehicle mileage awareness. In order to address the discrepancy between the vector space of mileage and that of the project, we put forth a mileage representation method for the maintenance demand prediction task. To capture the significant impact of key mileage and projects on future demand, we propose a learning module for key temporal information using a fusion of Long Short-Term Memory (LSTM) networks and attention mechanism. Moreover, to integrate maintenance mileage and projects, we propose a fusion method based on a gated unit. The experimental results obtained from real datasets demonstrate that the proposed model exhibits a superior performance compared to existing methods.

MMDG-DTI: Drug–target interaction prediction via multimodal feature fusion and domain generalization

Recently, deep learning has become the essential methodology for Drug–Target Interaction (DTI) prediction. However, the existing learning-based representation methods embed the prior knowledge encapsulated by the training data and, as such, acquire redundant domain information irrelevant to the DTI prediction, compromising the generalization capability of a deep network to unfamiliar instances. To tackle this limitation, we propose a novel DTI prediction method using Multi-Modal feature fusion and Domain Generalization (MMDG-DTI). Specifically, we first use pre-trained Large Language Models (LLMs) to obtain generalized textual features covering nearly the whole biological text vocabulary, with the capacity to handle unseen samples and extract robust and discriminative features. In parallel, we propose a unified structural feature extractor based on a hybrid Graph Neural Network (GNN). The extractor improves the performance of MMDG-DTI by extracting features from another modality and discarding the representation of domain-specific prior substructures. Then, we integrate the complementary information of LLM and GNN features using an innovative classifier, while enhancing the generalization ability with the help of Domain Adversarial Training (DAT) and contrastive learning. Last, we propose a novel cross-domain evaluation protocol to enhance the predication fidelity in practical applications. The quantitative and visualization results obtained on several benchmarking datasets demonstrate that MMDG-DTI achieves state-of-the-art performance under both single-domain and cross-domain settings, confirming the superiority of the proposed method. The datasets and codes will be available on the project homepage: https://github.com/JU-HuaY/MMDG-DTI.

CWT-ViT: A time–frequency representation and vision transformer-based framework for automated robotic surgical skill assessment

Surgical skill assessment currently hinges on the manual observations of senior surgeons, and the assessment process is inherently time-consuming and subjective. Hence, there is a need to develop machine learning-based automated robotic surgical skill assessment. However, the existing machine learning-based works are only built in either the time domain or frequency domain but have never considered the investigation on the time–frequency domain. To fill the research gap, we explore the representation of the surgery motion data in the time–frequency domain. In this study, we propose a novel automated robotic surgical skill assessment framework called Continuous Wavelet Transform-Vision Transformer (CWT-ViT). We apply continuous wavelet transform, i.e., a time–frequency representation method, to convert robotic surgery kinematic data to synthesis images. Furthermore, by taking advantage of the prior knowledge of the da Vinci surgical system, we design a four branches-based architecture, each branch representing a robotic manipulator. We have conducted extensive experiments and achieved comparable results on the benchmark robotic surgical skill dataset JHU-ISI Gesture and Skill Assessment Working Set (JIGSAWS). Our proposed CWT-ViT framework has demonstrated the feasibility of applying time–frequency representation for automated robotic surgical skill assessment using kinematic data. The code is available at https://github.com/yiming95/CWT-ViT-Surgery.

Closed-Form Method for Unified Far-Field and Near-Field Localization Based on TDOA and FDOA Measurements

When the near-field and far-field information of a target is uncertain, it is necessary to choose a suitable localization method. The modified polar representation (MPR) method integrates the two scenarios and achieves a unified localization with direction of arrival (DOA) estimation in the far field and position estimation in the near field. Previous studies have only proposed solutions for stationary environments and have not considered the motion factor. Therefore, this paper proposes a new unified positioning algorithm using multi-sensor time difference of arrival (TDOA) and frequency difference of arrival (FDOA) measurements without prior target source information. The method represents the position of the target source using MPR and describes the localization problem as a weighted least squares (WLS) problem with two constraints. We first obtain the initial estimates by WLS without considering the constraints and then investigate a two-step error correction method based on the constraints. The first step corrects the initial estimate using the Taylor series expansion technique, and the second step corrects the DOA estimate in the previous step using the direct error compensation technique based on the properties of the second constraint. Simulation experiments show that the method is effective for the unified positioning of moving targets and can achieve the Cramer–Rao lower bound (CRLB).

ProcessCarbonAgent: A large language models-empowered autonomous agent for decision-making in manufacturing carbon emission management

Knowledge-intensive production represents a primary trend in industrial manufacturing, which heavily relies on the production logs of large-scale, historically similar orders for enhancing production efficiency and process quality. These logs are essential for predicting resource allocation and identifying bottlenecks in throughput. As a result, root cause analysis of the production process state is crucial for supporting decision-making in these settings. However, current methodologies heavily depend on expert knowledge, making the analysis time-consuming and inefficient for large-scale, multivariable processes. Although the development of large language models and autonomous agents presents a potential solution, these models are limited in their direct interaction with event logs due to inadequate data representation, token constraints, and insufficient accuracy. Therefore, enabling the interactive capabilities of large language models to overcome these specific limitations in process event data and industrial domain illusions poses a significant challenge. To address these issues, this paper introduces the ProcessCarbonAgent framework, an autonomous agent empowered by large language models, designed to enhance decision-making within industrial processes. Initially, a process data agent combines predefined semantic text representation methods with process template prompting strategies to improve interaction capabilities. Subsequently, an intention agent utilizing self-information and large language models is developed to address context length limitations by identifying and eliminating redundancies. Finally, a two-stage confidence estimation method is implemented to refine the precision of decision-making assistance, thereby improving the accuracy of decisions supported by large language models. Experiments with textile industry carbon emission data reveal that the assisted decision-making scores employing a compression ratio of 0.5, closely align with scores from manually labeled evaluations, with a 98% overlap across scoring intervals. Moreover, in contrast to relying solely on the original evaluation method, the two-stage confidence estimation method has led to a 20% increase in accuracy performance. The ProcessCarbonAgent achieved scores of 16.64, 55.13, 26.32, and 34.17 on METEOR, BERTScore, NUBIA, and BLEURT, respectively. The results demonstrate that the ProcessCarbonAgent framework significantly enhances the decision-making process for high-carbon emission states in industrial production, providing technical support for the low-carbon transformation and intelligent upgrading of these processes.

Cartographic Generalization of Islands Using Remote Sensing Images for Multiscale Representation

The multi-scale representation of remote sensing images provides various levels of image information crucial for decision-making in GIS applications and plays a significant role in information processing, data analysis, and geographic modeling. Traditional methods for multi-scale representation of remote sensing images often struggle to simplify local details of individual targets while preserving the overall characteristics of target groups. These methods also encounter issues such as transitional texture distortion and rough final boundaries. This paper proposes a novel multi-scale representation method for remote sensing images based on computer vision techniques, which effectively maintains the overall characteristics of target groups. Initially, the K-means algorithm is employed to distinguish between islands and oceans. Subsequently, a superpixel segmentation algorithm is used to aggregate island groups and simplify the generated boundaries. Finally, texture synthesis and transfer are applied based on the original image to produce the aggregated island images. Evaluation metrics demonstrate that this method can generate multi-scale aggregated images of islands, effectively eliminate redundant information, and produce smooth boundaries.

MLC-DKT: A multi-layer context-aware deep knowledge tracing model

As a fundamental research task for intelligent education, knowledge tracing (KT) aims to trace learners’ dynamic knowledge states and predict their future performance based on their historical response data. However, learners’ cognitive processes cannot be studied in isolation from the context; nevertheless, existing knowledge tracing models mainly use knowledge topology as auxiliary information but do not deeply explore the contextual information of exercises. Thus, an in-depth exploration of multi-layer contextual information in learning scenarios is required, focusing on the cognitive differences among different learners. Furthermore, knowledge tracing models exhibit limited interpretability compared with cognition diagnosis methods integrating educational priors. To address these issues, we propose a multi-layer context-aware deep knowledge tracing (MLC-DKT) model. Specifically, we first present the multi-layer contextual representation method involving knowledge concepts and exercises. In the knowledge state tracing module, we incorporate the learners’ knowledge priors into the attention mechanism to capture the evolution of the learners’ knowledge. Then, we develop a calculation method for the similarity effect among the contextual information, which helps estimate the learners’ knowledge states. In the learning performance prediction module, we introduce the guessing and slipping factors from the cognition diagnosis model to optimize the MLC-DKT model, further improving the performance of the KT model. Finally, we conduct extensive experiments on real-world datasets, demonstrating that the MLC-DKT model provides improved predictions of learner performance. Moreover, the MLC-DKT model realizes contextual awareness from knowledge concepts and exercises, making the predictions more interpretable.

Anomalous behavior detection based on optimized graph embedding representation in social networks

Anomalous behaviors in social networks can lead to privacy leaks and the spread of false information. In this paper, we propose an anomalous behavior detection method based on optimized graph embedding representation. Specifically, the user behavior logs are first extracted into a social network user behavior temporal knowledge graph, based on which the graph embedding representation method is used to transform the network topology and temporal information in the user behavior knowledge graph into structural embedding vectors and temporal information embedding vectors, and the hybrid attention mechanism is used to merge the two types of vectors to obtain the final entity embedding to complete the prediction and complementation of the temporal knowledge graph of user behavior. We use graph neural networks, which use the temporal information of user behaviors as a time constraint and capture both user behavioral and semantic information. It converts the two parts of information into vectors for concatenation and linear transformation to obtain a comprehensive representation vector of the whole subgraph, as well as joint deep learning models to evaluate abnormal behavior. Finally, we perform experiments on the Yelp dataset to validate that our method achieves a 9.56% improvement in the F1-score.

A novel density‐based representation for point cloud and its ability to facilitate classification

AbstractCurrently, in the field of processing 3D point cloud data, two primary representation methods have emerged: point‐based methods and voxel‐based methods. However, the former suffer from significant computational costs and lack the ease of handling exhibited by voxel‐based methods. Conversely, the later often encounter challenges related to information loss resulting from downsampling operations, thereby impeding subsequent tasks. To address these limitations, this article introduces a novel density‐based representation method for voxel partitioning. Additionally, a corresponding network structure is devised to extract features from this specific density representation, thereby facilitating the successful completion of classification tasks. The experiments are implemented on ModelNet40 and MNIST demonstrate that the proposed 3D convolution can achieve the‐state‐of‐the‐art performance based on the voxels.