Related Datasets Research Articles

Improving semantic segmentation with knowledge reasoning network

Most methods cannot segment the semantic regions accurately due to the lack of global-level supervision or guidance of external knowledge. To overcome this limitation, we propose a Knowledge Reasoning Network (KRNet) that consists of two crucial modules: (1) a prior knowledge mapping module that incorporates external knowledge by graph convolutional network to guide learning semantic representations and (2) a knowledge reasoning module that correlates these representations with a graph built on the external knowledge and explores their interactions via the knowledge reasoning. In the prior knowledge mapping module, we adopt an algorithm to mine knowledge from an external large-scale relational modeling dataset. In the knowledge reasoning module, we adopt an iterative mechanism to perform knowledge reasoning and explore the interaction between features. Reasoning makes the spatial distribution of categories more significant. We establish state-of-the-art results on Cityscapes and ADE20K datasets, which demonstrates the effectiveness of our methods on semantic segmentation.

Infrared pedestrian detection using improved UNet and YOLO through sharing visible light domain information

The detection of pedestrians using infrared images is essential for security during automatic driving and monitoring under low illumination or occlusion conditions. However, detecting pedestrians in infrared images can be challenging due to the lack of clear visual cues and background noise. To address these challenges, we proposed an efficient approach based on an improved UNet and YOLO network that shares visible light information from multiple related datasets. The proposed approach achieves high detection accuracies, with mean average precision values of 87.7%, 88%, and 97.5% on three different datasets, FLIR, M3FD, and LLVIP, respectively. Moreover, the designed infrared pedestrian detection network is deployed on an edge computing device and tested with a dual-light handheld camera, achieving a real-time detection speed of 25.6 FPS with a superior trade-off between speed and accuracy. These results demonstrate the effectiveness of the proposed approach in detecting pedestrians in challenging infrared image conditions.

Identification and validation of potential hypoxia-related genes associated with coronary artery disease.

Introduction: Coronary artery disease (CAD) is one of the most life-threatening cardiovascular emergencies with high mortality and morbidity. Increasing evidence has demonstrated that the degree of hypoxia is closely associated with the development and survival outcomes of CAD patients. However, the role of hypoxia in CAD has not been elucidated. Methods: Based on the GSE113079 microarray dataset and the hypoxia-associated gene collection, differential analysis, machine learning, and validation of the screened hub genes were carried out. Results: In this study, 54 differentially expressed hypoxia-related genes (DE-HRGs), and then 4 hub signature genes (ADM, PPFIA4, FAM162A, and TPBG) were identified based on microarray datasets GSE113079 which including of 93 CAD patients and 48 healthy controls and hypoxia-related gene set. Then, 4 hub genes were also validated in other three CAD related microarray datasets. Through GO and KEGG pathway enrichment analyses, we found three upregulated hub genes (ADM, PPFIA4, TPBG) were strongly correlated with differentially expressed metabolic genes and all the 4 hub genes were mainly enriched in many immune-related biological processes and pathways in CAD. Additionally, 10 immune cell types were found significantly different between the CAD and control groups, especially CD8 T cells, which were apparently essential in cardiovascular disease by immune cell infiltration analysis. Furthermore, we compared the expression of 4 hub genes in 15 cell subtypes in CAD coronary lesions and found that ADM, FAM162A and TPBG were all expressed at higher levels in endothelial cells by single-cell sequencing analysis. Discussion: The study identified four hypoxia genes associated with coronary heart disease. The findings provide more insights into the hypoxia landscape and, potentially, the therapeutic targets of CAD.

Employing Molecular Conformations for Ligand-Based Virtual Screening with Equivariant Graph Neural Network and Deep Multiple Instance Learning.

Ligand-based virtual screening (LBVS) is a promising approach for rapid and low-cost screening of potentially bioactive molecules in the early stage of drug discovery. Compared with traditional similarity-based machine learning methods, deep learning frameworks for LBVS can more effectively extract high-order molecule structure representations from molecular fingerprints or structures. However, the 3D conformation of a molecule largely influences its bioactivity and physical properties, and has rarely been considered in previous deep learning-based LBVS methods. Moreover, the relative bioactivity benchmark dataset is still lacking. To address these issues, we introduce a novel end-to-end deep learning architecture trained from molecular conformers for LBVS. We first extracted molecule conformers from multiple public molecular bioactivity data and consolidated them into a large-scale bioactivity benchmark dataset, which totally includes millions of endpoints and molecules corresponding to 954 targets. Then, we devised a deep learning-based LBVS called EquiVS to learn molecule representations from conformers for bioactivity prediction. Specifically, graph convolutional network (GCN) and equivariant graph neural network (EGNN) are sequentially stacked to learn high-order molecule-level and conformer-level representations, followed with attention-based deep multiple-instance learning (MIL) to aggregate these representations and then predict the potential bioactivity for the query molecule on a given target. We conducted various experiments to validate the data quality of our benchmark dataset, and confirmed EquiVS achieved better performance compared with 10 traditional machine learning or deep learning-based LBVS methods. Further ablation studies demonstrate the significant contribution of molecular conformation for bioactivity prediction, as well as the reasonability and non-redundancy of deep learning architecture in EquiVS. Finally, a model interpretation case study on CDK2 shows the potential of EquiVS in optimal conformer discovery. The overall study shows that our proposed benchmark dataset and EquiVS method have promising prospects in virtual screening applications.

Similarity learning of product descriptions and images using multimodal neural networks

Multimodal deep learning is an emerging research topic in machine learning and involves the parallel processing of different modalities of data such as texts, images and audiovisual data. Well-known application areas are multimodal image and video processing as well as speech recognition. In this paper, we propose a multimodal neural network that measures the similarity of text-written product descriptions and images and has applications in inventory reconciliation and search engine optimization. We develop two models. The first takes image and text data, each processed by convolutional neural networks, and combines the two modalities. The second is based on a bidirectional triplet loss function. We conduct experiments using ABO! (ABO!) dataset and an industry-related dataset used for the inventory reconciliation of a mechanical engineering company. Our first model achieves an accuracy of 92.37% with ResNet152 on the ABO! dataset and 99.11% with MobileNetV3_Large on our industry-related dataset. By extending this model to a model with three inputs, two text inputs and one image input, we greatly improve the performance and achieve an accuracy of 97.57% on the ABO! dataset and 99.83% with our industry related inventory dataset. Our second model based on the triplet loss achieves only an accuracy of 73.85% on the ABO! dataset. However, our experiments demonstrate that multimodal networks consistently perform better when measuring the similarity of products, even in situations where one modality lacks sufficient data, because it is complemented with the other modality. Our proposed approaches open up several possibilities for further optimization of search engines.

Quantifying the competitiveness of a dataset in relation to general preferences

Quantifying the competitiveness of a dataset in relation to general preferences

Zależności Z-R dla różnych typów opadów jako narzędzie do radarowego szacowania wielkości opadów = The Z-R relationships for different types of precipitation as a tool for radar-based precipitation estimation

An alternative to the use of rain gauges as sources of precipitation data is provided by laser disdrometers, which inter alia allow for high-temporal-resolution measurement of the reflectivity (Z) and intensity (R) of precipitation. In the study detailed here, an OTT Parsivel1 laser disdrometer located at the Meteorological Station of Warsaw University of Life Sciences (SGGW) generated the 95,459 Z-R data pairs recorded across 1-min time intervals that were subject to further study. Included values for the reflectivity and instantaneous intensity of precipitation were found to be in the respective ranges of -9.998‑67.898 dBZ and 0.004‑153.519 mm h−1 (given that values for precipitation intensity below 0.004 mm h−1 were excluded from further consideration). The material obtained covered the months from April to October in the years 2012‑2014 and 2019‑2020 (30 months in total), which were selected for the study due to the completeness of data. The measured reflectivity and intensity data for precipitation were used to establish the relationship pertaining between the two (by reference to descriptive parameters a and b), with such results considered to contribute to the improved calibration of meteorological radars, and hence to more-accurate radar-based estimates of amounts of precipitation. The Z-R relationship as determined for all measurement data offered a first step in the research process, whose core objective was nevertheless to determine separate Z-R relationships for datasets on rain, rain with snow (sleet), and snow (given that precipitation in the form of hail did not occur during the surveyed measurement periods). That said, it is important to note that only a few Polish studies have in any way involved disdrometer-based measurement of precipitation reflectivity and intensity, as well as the relationships between these aspects. In the event, the Z-R relationships obtained for the measurement sets were characterised by high values for coefficients of correlation (in the range 0.96‑0.97) and determination, as well as low values for the root mean-square error (ranging from 0.29 to 0.34). Statistics point to a good fit of the Z-R relationships (regression lines) to the specified datasets. Values noted for parameter a (the multiplier in the power-type Z-R relationship) were seen to differ significantly in relation to rain, rain with snow, and snow, being of 285.56, 76.07 and 914.74 respectively. In contrast, values noted for parameter b (the exponent) varied only across the narrow range of 1.47‑1.62. The obtained research results for parameter a indicate the need to consider Z-R relationships matched to specific types of precipitation in the data processing procedure of radar data. This could increase the accuracy of estimating precipitation amounts using radars belonging to the nationwide POLRAD system. The relationships Z = 285.56R1.47 for rainfall (as the dataset’s dominant type of precipitation), as well as Z = 293.76R1.46 for all data, proved highly similar to the classic relationship obtained for convective rainfall by Hunter (1996), as given by Z = 300R1.4. On the other hand, the values of the a parameter in the Z-R relationships fond for the two datasets proved to be much larger than those in the model developed by Marshall and Palmer (1948), which took the form Z = 200R1,6 and has been the relationship used in Poland as radar images are created.

Relation Extraction Based on Prompt Information and Feature Reuse

ABSTRACT To alleviate the problem of under-utilization features of sentence-level relation extraction, which leads to insufficient performance of the pre-trained language model and underutilization of the feature vector, a sentence-level relation extraction method based on adding prompt information and feature reuse is proposed. At first, in addition to the pair of nominals and sentence information, a piece of prompt information is added, and the overall feature information consists of sentence information, entity pair information, and prompt information, and then the features are encoded by the pre-trained language model ROBERTA. Moreover, in the pre-trained language model, BIGRU is also introduced in the composition of the neural network to extract information, and the feature information is passed through the neural network to form several sets of feature vectors. After that, these feature vectors are reused in different combinations to form multiple outputs, and the outputs are aggregated using ensemble-learning soft voting to perform relation classification. In addition to this, the sum of cross-entropy, KL divergence, and negative log-likelihood loss is used as the final loss function in this paper. In the comparison experiments, the model based on adding prompt information and feature reuse achieved higher results of the SemEval-2010 task 8 relational dataset.

Resveratrol drives ferroptosis of acute myeloid leukemia cells through Hsa-miR-335-5p/NFS1/ GPX4 pathway in a ROS-dependent manner.

To explore the potential target to induce ferroptosis for treating acute myeloid leukemia (AML) as well as its mechanism and latent drugs. Using the keyword "acute myelogenous leukemia", the related dataset in TCGA and GEO were used for searching differentially expressed genes. After the filtrate by ROC curve, AUC values, and survival analysis, RT-qPCR as well as Western-blot analysis were performed to verify the high expression level of NFS1 in AML-193 and OCI-AML-3 cells. After CCK-8 detection with and without various cell death inhibitors, ferroptosis were further detected by the expression level of GPX4. After taking the intersection in Starbase and TargetScan, the upstream regulatory miRNA of NFS1 was found. Then the relation of hsa-miR-335-5p, NFS1, as well as GPX4, was ascertained by knockdown and overexpression study in AML-193 and OCI-AML-3 cells. In addition, cellular ROS was detected by DCFH-DA. Finally, resveratrol was used to intensify ferroptosis of AML-193 and OCI-AML-3 cells. NFS1 was highly expressed in AML cells, positively associated with AML-related mortality, and can be used to diagnose AML. Knockout of NFS1 facilitated ROS accumulation and ferroptosis-associated labile iron pool increase. si-NFS1 can inhibit the expression level of GPX4, facilitate ROS accumulation and induce ferroptosis-associated labile iron pool increase. Besides, overexpressed GPX4 can lead to down-regulated cell death after si-NFS1 treatment. Hsa-miR-335-5p was found as the upstream regulator of NFS1. The expression of NFS1 can be up-regulated by sh-hsa-miR-335-5p transfection and can be inhibited by hsa-miR-335-5p transfection. Resveratrol was found can increase the expression level of hsa-miR-335-5p and decrease the expression of NFS1 and GPX4. Resveratrol can intensify ferroptosis of AML cells via Hsa-miR-335-5p/NFS1/ GPX4 pathway through a ROS-dependent manner.

Artificial Intelligence and Mathematical Models of Power Grids Driven by Renewable Energy Sources: A Survey

To face the impact of climate change in all dimensions of our society in the near future, the European Union (EU) has established an ambitious target. Until 2050, the share of renewable power shall increase up to 75% of all power injected into nowadays’ power grids. While being clean and having become significantly cheaper, renewable energy sources (RES) still present an important disadvantage compared to conventional sources. They show strong fluctuations, which introduce significant uncertainties when predicting the global power outcome and confound the causes and mechanisms underlying the phenomena in the grid, such as blackouts, extreme events, and amplitude death. To properly understand the nature of these fluctuations and model them is one of the key challenges in future energy research worldwide. This review collects some of the most important and recent approaches to model and assess the behavior of power grids driven by renewable energy sources. The goal of this survey is to draw a map to facilitate the different stakeholders and power grid researchers to navigate through some of the most recent advances in this field. We present some of the main research questions underlying power grid functioning and monitoring, as well as the main modeling approaches. These models can be classified as AI- or mathematically inspired models and include dynamical systems, Bayesian inference, stochastic differential equations, machine learning methods, deep learning, reinforcement learning, and reservoir computing. The content is aimed at the broad audience potentially interested in this topic, including academic researchers, engineers, public policy, and decision-makers. Additionally, we also provide an overview of the main repositories and open sources of power grid data and related data sets, including wind speed measurements and other geophysical data.

BreastDM: A DCE-MRI dataset for breast tumor image segmentation and classification

Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) has shown high sensitivity to diagnose breast cancer. However, few computer-aided algorithms focus on employing DCE-MR images for breast cancer diagnosis due to the lack of publicly available DCE-MRI datasets. To address this issue, our work releases a new DCE-MRI dataset called BreastDM for breast tumor segmentation and classification. In particular, a dataset of 232 patients selected with DCE-MR images for benign and malignant cases is established. Each case consists of three types of sequences: pre-contrast, post-contrast, and subtraction sequences. To show the difficulty of breast DCE-MRI tumor image segmentation and classification tasks, benchmarks are achieved by state-of-the-art image segmentation and classification algorithms, including conventional hand-crafted based methods and recently-emerged deep learning-based methods. More importantly, a local–global cross attention fusion network (LG-CAFN) is proposed to further improve the performance of breast tumor images classification. Specifically, LG-CAFN achieved the highest accuracy (88.20%, 83.93%) and AUC value (0.9154,0.8826) in both groups of experiments. Extensive experiments are conducted to present strong baselines based on various typical image segmentation and classification algorithms. Experiment results also demonstrate the superiority of the proposed LG-CAFN to other breast tumor images classification methods. The related dataset and evaluation codes are publicly available at smallboy-code/Breast-cancer-dataset.

ECG Heartbeat classification using deep transfer learning with Convolutional Neural Network and STFT technique

Electrocardiogram (ECG) is a simple non-invasive measure to identify heart-related issues such as irregular heartbeats known as arrhythmias. While artificial intelligence and machine learning is being utilized in a wide range of healthcare related applications and datasets, many arrhythmia classifiers using deep learning methods have been proposed in recent years. However, sizes of the available datasets from which to build and assess machine learning models is often very small and the lack of well-annotated public ECG datasets is evident. In this paper, we propose a deep transfer learning framework that is aimed to perform classification on a small size training dataset. The proposed method is to fine-tune a general-purpose image classifier ResNet-18 with MIT-BIH arrhythmia dataset in accordance with the AAMI EC57 standard. This paper further investigates many existing deep learning models that have failed to avoid data leakage against AAMI recommendations. We compare how different data split methods impact the model performance. This comparison study implies that future work in arrhythmia classification should follow the AAMI EC57 standard when using any including MIT-BIH arrhythmia dataset.

Social learning and the complex contagion of political symbols in Twitter: The case of the yellow ribbon in Catalonia

In this paper, we empirically test whether the spread of political symbols in Twitter is due to complex contagion. We analyzed behavior consisting of editing the Twitter account name to include an icon with a yellow ribbon; a symbol that represents the demand for the release of imprisoned Catalan politicians and civil leaders. To test this hypothesis, we used a behavioral, non-reactive, relational, and dynamic dataset of a large sample of potential users. First, we show that the probability of displaying a ribbon is associated with the proportion of peers who also display it (friends that share their support for the political cause). Second, we rule out alternative explanations as simple contagion and homophily. To rule out simple contagion, we run three empirically calibrated, agent-based simulations. We use our dataset to rule out homophily. And third, we suggest that adoption cannot be interpreted as the result of a compliance mechanism or as the result of normative pressures. Instead, the most plausible micro-level generative mechanism that leads to a complex contagion pattern is a peer learning process. Our study makes several contributions to the field. We show how digital data can be effectively used to identify new explananda and test the plausibility of competing behavioral explanatory mechanisms. We also contribute to the development of the theory of complex contagions. Our study widens the set of conditions for complex contagion and the set of reasons to explain why complex contagion might occur.

Abnormal discharge detection using adaptive neuro-fuzzy inference method with probability density-based feature and modified subtractive clustering

Abnormal discharge (AD) is a discharge mode in electroencephalogram (EEG) with sharp outlines. Lack of related open-access datasets and insufficient annotation hamper the development of AD detection, while cognitive neuroscience, clinical research and pilots’ neural screening demand stable and reliable AD detection methods. An adaptive neuro-fuzzy inference method for AD detection is proposed in this work. First, a probability-density-based (PD) method is proposed to extract lognormal amplitude features from EEG envelopes. Second, the subtractive clustering method (SCM) is modified so that clustering radii can adapt to cluster shapes for each dimension instead of using identical radii for each cluster. The outputs of modified SCM (mSCM), coordinates of cluster centers and adjusting rates for radius components are used to automatically initialize Gaussian membership functions of adaptive-network-based fuzzy inference system (ANFIS). Finally, we conducted multiple experiments to validate mSCM-based ANFIS, comparing it with traditional machine learning classifiers (support vector machines, multi-layer perceptrons, decision trees, and random forests) and the state-of-the-art deep learning-based time-series classification methods InceptionTime and Minirocket, using synthetic data, small-size datasets and a private EEG dataset. Results show that combining PD features with features proposed in previous studies, such as smoothed nonlinear energy operator features and discrete wavelet transform features, achieved higher accuracy (95.13 ± 0.86%) and recall (89.17 ± 3.06%) than other feature combinations. mSCM created more suitable cluster boundaries than SCM on small-size datasets, and its clustering results demonstrated potential in helping interpretation of classification rules built in the ANFIS network. Results of comparison experiments showed that mSCM-based ANFIS produced competitive results in accuracy and recall for AD detection, with lower computational cost, compared to the top 2 results obtained by InceptionTime and Minrocket.

Insights into fuel behaviour during relatively fast thermal transients based on calculations for two tests of the Halden IFA-507 experiment

Outcomes of the project “Comprehensive Verification of the FALCON Code for Calculation of Nuclear Fuel Temperature” relating to calculation of fuel temperature during relatively fast thermal transients are presented. Good prediction capabilities of the FALCON MOD01 code coupled with the GRSW-A code are shown as applied to the data of the TF3 and TF5 tests from the Transient Temperature Experiment IFA-507. The IFA-507 related dataset of the OECD/NEA International Fuel Performance Experiments (IFPE) Database is extended by the reconstructed dynamics of the axial power distribution in the rods during the transient phase of the experiment. Based on the code calculation, the time constant of the thermal fuel response to a power transient is estimated.

A Question Answering System for Situation Puzzle with SPQA

There are many questions answering (QA) system built for solving QA tasks. In 2020 and 2022, Allen Institute and the University of Washington proposed UnifiedQA and UnifiedQA-v2. Their core concept is that the semantic understanding and reasoning capabilities required by models are common, and may not require format specific models although the QA task forms are different. Behind this concept, I build a new QA model named SPQA, aiming to answer the situation puzzle questions by adding new situation-puzzle related dataset (SpQ). In addition, I evaluate the performance of SPQA and UnifiedQA-v2 for fine-tuning and prompt-tuning. The results of fine-tuning indicate that SpQ dataset is important for fine-tuning and prompt-tuning to answer situation puzzle questions well, but also make the answering ability of normal yes/no questions worse. Eventually, the results of prompt-tuning indicate that the effects of SpQ is larger and more significant on situation puzzle questions and normal yes/no questions under the same data scale. In the future work, the further research like building larger SpQ dataset should be considered.

Shaping influence in governance networks: The role of motivations and information exchange

AbstractIn governance networks, some actors might have more influence than others in the group's collective decision‐making. This paper investigates whether an actor's prosocial and/or self‐interested motivations to participate in a governance network help predict its level of influence in the group. We argue that information exchange is an important mediator in this relationship because an actor's tendency to actively diffuse information will depend on its motivations; while other participants being exposed to information from an actor are likely to increase the actor's influence on them. Using a unique relational dataset from 10 anti‐corruption multi‐stakeholder partnerships (MSPs) in Latin America, Africa and Eurasia, we find that self‐interested actors, rather than prosocially motivated ones, take the lead in information‐exchange activities. The data also shows how this central role in turn increases perceived influence of self‐interested actors among other participants, conditioning potentially the direction of agreed‐upon collective objectives.

Fostering Trustworthiness of Federated Learning Ecosystem through Realistic Scenarios

The present study thoroughly evaluates the most common blocking challenges faced by the federated learning (FL) ecosystem and analyzes existing state-of-the-art solutions. A system adaptation pipeline is designed to enable the integration of different AI-based tools in the FL system, while FL training is conducted under realistic conditions using a distributed hardware infrastructure. The suggested pipeline and FL system’s robustness are tested against challenges related to tool deployment, data heterogeneity, and privacy attacks for multiple tasks and data types. A representative set of AI-based tools and related datasets have been selected to cover several validation cases and distributed to each edge device to closely reflect real-world scenarios. The study presents significant outcomes of the experiments and analyzes the models’ performance under different realistic FL conditions, while highlighting potential limitations and issues that occurred during the FL process.

Chinese Document-Level Emergency Event Extraction Dataset and Corresponding Methods

Structured extraction of emergency event information can effectively enhance the ability to respond to emergency events. This article focuses on extracting Chinese document-level emergency events, which entails addressing two key issues in this field: related datasets and the problem of role overlapping between candidate entities, which has been overlooked in existing DEE (document-level event extraction) studies that predominantly employed sequence annotation for candidate entity extraction. To tackle these challenges, we constructed a Chinese document-level emergency extraction dataset (CDEEE) and provides annotations for argument scattering, multiple events, and role overlapping. Additionally, a model named RODEE is proposed to address the role overlapping problem in DEE tasks. RODEE employs two independent modules to represent the head and tail positions of candidate entities, and utilizes a multiplication attention mechanism to interact between the two, generating a scoring matrix. Subsequently, role-overlapping candidate entities are predicted to facilitate the completion of DEE tasks. Experiments were conducted on our manually annotated dataset, CDEEE, and the results show that RODEE effectively solves the problem of role overlapping among candidate entities, resulting in improved performance of the DEE model, with an F1 value of 77.7%.

Efficient Sensor Scheduling Strategy Based on Spatio-Temporal Scope Information Model

In this paper, based on the information entropy and spatio-temporal correlation of sensing nodes in the Internet of Things (IoT), a Spatio-temporal Scope Information Model (SSIM) is proposed to quantify the scope of the valuable information of sensor data. Specifically, the valuable information of sensor data decays with space and time, which can be used to guide the system to make efficient sensor activation scheduling decisions for regional sensing accuracy. A simple sensing and monitoring system with three sensor nodes is investigated in this paper, and a single-step scheduling decision mechanism is proposed for the optimization problem of maximizing valuable information acquisition and efficient sensor activation scheduling in the sensed region. Regarding the above mechanism, the scheduling results and approximate numerical bounds on the node layout between different scheduling results are obtained through theoretical analyses, which are consistent with simulation. In addition, a long-term decision mechanism is also proposed for the aforementioned optimization issues, where the scheduling results with different node layouts are derived by modeling as a Markov decision process and utilizing the Q-learning algorithm. Concerning the above two mechanisms, the performance of both is verified by conducting experiments using the relative humidity dataset; furthermore, the differences in performance and limitations of the model are discussed and summarized.