Identification Performance Research Articles

Accurate prediction of protein–ligand interactions by combining physical energy functions and graph-neural networks

We introduce an advanced model for predicting protein–ligand interactions. Our approach combines the strengths of graph neural networks with physics-based scoring methods. Existing structure-based machine-learning models for protein–ligand binding prediction often fall short in practical virtual screening scenarios, hindered by the intricacies of binding poses, the chemical diversity of drug-like molecules, and the scarcity of crystallographic data for protein–ligand complexes. To overcome the limitations of existing machine learning-based prediction models, we propose a novel approach that fuses three independent neural network models. One classification model is designed to perform binary prediction of a given protein–ligand complex pose. The other two regression models are trained to predict the binding affinity and root-mean-square deviation of a ligand conformation from an input complex structure. We trained the model to account for both deviations in experimental and predicted binding affinities and pose prediction uncertainties. By effectively integrating the outputs of the triplet neural networks with a physics-based scoring function, our model showed a significantly improved performance in hit identification. The benchmark results with three independent decoy sets demonstrate that our model outperformed existing models in forward screening. Our model achieved top 1% enrichment factors of 32.7 and 23.1 with the CASF2016 and DUD-E benchmark sets, respectively. The benchmark results using the LIT-PCBA set further confirmed its higher average enrichment factors, emphasizing the model’s efficiency and generalizability. The model’s efficiency was further validated by identifying 23 active compounds from 63 candidates in experimental screening for autotaxin inhibitors, demonstrating its practical applicability in hit discovery.Scientific contributionOur work introduces a novel training strategy for a protein–ligand binding affinity prediction model by integrating the outputs of three independent sub-models and utilizing expertly crafted decoy sets. The model showcases exceptional performance across multiple benchmarks. The high enrichment factors in the LIT-PCBA benchmark demonstrate its potential to accelerate hit discovery.

Surface Defect Identification of Strip Steel Using ViT‐RepVGG

In the production of strip steel, surface defect identification is crucial for improving product quality and ensuring smooth subsequent processes. Existing technologies face challenges such as low detection efficiency and susceptibility to environmental noise. This article employs an automated deep learning method without requiring consideration of complex environmental changes and proposes an improved RepVGG (ViT‐RepVGG) model for surface defect identification. The model is based on the RepVGG architecture, and the study explores the impact of incorporating the self‐attention mechanism of ViT under various addition strategies on model performance. A comparison is made between the optimized model and classic network models, as well as recently published models, in terms of identification performance. The research also examines the performance variations of the model under different hyperparameter settings and its identification performance for six types of defects. The results indicate that adding the ViT module to stage 3 of the A1‐type RepVGG, with a learning rate, optimizer, and activation function set to 0.0001, Adam, and Gelu, respectively, yields the optimal ViT‐RepVGG model performance. These findings demonstrate the feasibility of enhancing classification performance by incorporating the self‐attention mechanism into neural networks, providing an effective foundation for the online identification of strip steel surface defects.

MEF-AlloSite: an accurate and robust Multimodel Ensemble Feature selection for the Allosteric Site identification model

A crucial mechanism for controlling the actions of proteins is allostery. Allosteric modulators have the potential to provide many benefits compared to orthosteric ligands, such as increased selectivity and saturability of their effect. The identification of new allosteric sites presents prospects for the creation of innovative medications and enhances our comprehension of fundamental biological mechanisms. Allosteric sites are increasingly found in different protein families through various techniques, such as machine learning applications, which opens up possibilities for creating completely novel medications with a diverse variety of chemical structures. Machine learning methods, such as PASSer, exhibit limited efficacy in accurately finding allosteric binding sites when relying solely on 3D structural information.Scientific ContributionPrior to conducting feature selection for allosteric binding site identification, integration of supporting amino-acid–based information to 3D structural knowledge is advantageous. This approach can enhance performance by ensuring accuracy and robustness. Therefore, we have developed an accurate and robust model called Multimodel Ensemble Feature Selection for Allosteric Site Identification (MEF-AlloSite) after collecting 9460 relevant and diverse features from the literature to characterise pockets. The model employs an accurate and robust multimodal feature selection technique for the small training set size of only 90 proteins to improve predictive performance. This state-of-the-art technique increased the performance in allosteric binding site identification by selecting promising features from 9460 features. Also, the relationship between selected features and allosteric binding sites enlightened the understanding of complex allostery for proteins by analysing selected features. MEF-AlloSite and state-of-the-art allosteric site identification methods such as PASSer2.0 and PASSerRank have been tested on three test cases 51 times with a different split of the training set. The Student’s t test and Cohen’s D value have been used to evaluate the average precision and ROC AUC score distribution. On three test cases, most of the p-values (<0.05\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$< 0.05$$\\end{document}) and the majority of Cohen’s D values (>0.5\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$> 0.5$$\\end{document}) showed that MEF-AlloSite’s 1–6% higher mean of average precision and ROC AUC than state-of-the-art allosteric site identification methods are statistically significant.

An efficient enhanced stacked auto encoder assisted optimized deep neural network for forecasting Dry Eye Disease

Meibomian Gland Dysfunction (MGD) and Dry Eye Disease (DED) comprise two of the most significant eye diseases, impacting millions of sufferers worldwide. Several etiological factors influence the early symptoms of DED. Early diagnosis and treatment of erectile dysfunction may significantly improve the Quality of Life (QoL) for people. The current study introduces the ESAE-ODNN, an improved stacked autoencoder-aided optimised deep neural network, as a new way to predict DED using feature selection (FS), feature extraction (FE), and classification. The approach described here is novel because it merges chaotic maps into FS, employs SLSTM-STSA for improved classification accuracy (CA), and optimizes with the adaptive quantum rotation of the Enhanced Quantum Bacterial Foraging Optimisation Algorithm (EQBFOA). The present study enhances prediction functions by extracting MGD-related features and complicated relationships from the DED dataset. To ensure essential feature identification, the ESAE minimizes irrelevant and redundant features. To predict the DED, the ESAE first applies FE and then implements an ODNN classifier. This method fine-tunes the ODNN framework to enhance the effectiveness of the classification. The proposed ESAE-ODNN classification system efficiently assists in the early diagnosis of DED. Combining advanced Deep Learning (DL) methods with optimization can help us understand MGD features better and sort the data with the best accuracy (96.34%). The experimental evaluation with relevant performance metrics indicates that the proposed method is efficient in diverse aspects: accurate identification, reduced complexity, and fine-tuned performance. The ESAE-ODNN’s robustness in handling intricate feature indications and high-dimensional data outperforms the existing state-of-the-art techniques.

Multi-year comparison of VITEK MS performance for identification of rarely encountered pathogenic Gram-negative organisms (GNOs) in a large integrated Canadian healthcare region.

This multi-year study (2014-2019) compared identification of rare and unusual Gram-negative organisms (GNOs) by matrix-assisted laser desorption ionization-time of flight mass spectrometry (MALDI-TOF MS) (VITEK MS, bioMérieux, Laval Que.) to 16S rRNA gene sequencing (16S) according to our laboratories routine workflow; 16S is done if initial MALDI-TOF MS gave discordant, wrong, or no results. GNB isolates were first analyzed by standard phenotypic methods and MALDI-TOF MS using direct deposit-full formic acid extraction; proteomics was repeated if no result occurred. Medically approved 16S analyses were done using fast protocols. Isolate sequences were analyzed using the Integrated Database Network System (IDNS3) bacterial database (SmartGene, Lausanne, Switzerland). Three hundred thirty-one GNOs including 251 (76%) aerobic Gram-negative bacilli (GNB), 63 (19%) fastidious Gram-negative coccobacilli (fGNCBs), and 17 (5%) Campylobacterales (CAMPB) isolates were recovered from 304 specimens; >1 isolate was recovered from 19 (6%). GNOs were mainly recovered from blood cultures (31.6%) and lower respiratory specimens (43%) (one-half were isolated from cystic fibrosis patients). Accurate genus vs species identities were obtained for 67.7% and 32.5% aerobic GNBs, 73% and 60% fGNCBs, and 23.5% CAMPB (with no discrepant species), respectively. Wrong or no results were obtained for 81 (32.3%) aerobic GNBs, 17 (27%) fGNCBs, and 13 (72.2%) CAMPB. No results or misidentifications occurred for 33% of aerobic GNBs, 26% of fGNCBs, and 76.5% of CAMPB due to absence of species in the instrument's database. VITEK MS performance remained stable for aerobic GNBs and fGNCBs but improved for CAMPB with addition of Campylobacter rectus and Campylobacter curvus to the database. 16S remains important for identification of GNOs when proteomics fails.IMPORTANCEMatrix-assisted laser desorption ionization-time of flight mass spectrometry (MALDI-TOF MS) has transformed the identification of commonly encountered Gram-negative organisms (GNOs) in the clinical laboratory, but rare and unusual organisms continue to challenge the technology. This study verified performance of VITEK MS for identification of a broad range of rare and unusual clinical GNO isolates by our large reference laboratory workflow over a multi-year period. Although most GNOs were accurately identified by MALDI-TOF MS, a small number of clinical isolates (~1%-6%) required 16S sequencing for identification depending on the GNO category. Approximately one-third of aerobic Gram-negative bacilli (GNBs) and two-thirds of Campylobacterales could not be accurately identified by proteomics due to lack of an organism in the instrument's database. MALDI-TOF MS databases should be continuously updated and validated, and laboratories should have a workflow for identification of unusual or rarely encountered aerobic, fastidious, and Campylobacterales GNOs that includes 16S rRNA gene sequencing whenever proteomics cannot give a definitive identification.

Fast Iterative Sample Transfer Identification Method for Dynamic Systems Under Non‐identical Distribution

ABSTRACTThis paper proposes a method to improve the identification performance of linear dynamic systems by utilizing knowledge from samples of non‐identical distribution systems. Traditional identification methods heavily rely on the quality of the dataset, such as sample length and noise level, which constrains their performance due to the assumption of identical distribution. Motivated by the concept of sample‐based transfer learning, we propose a sample transfer identification method and derive the condition to avoid negative transfer. We develop a fast iterative transfer identification method for low storage costs, considering the computational burden imposed by the sample size from the source system. Additionally, based on the fast iterative transfer identification method, considering the need to update the current measurement data model in real time, a fast iterative online sample transfer identification method is explored. Through simulations, we validate the effectiveness and superiority of the proposed methods. The results show that sample transfer identification is superior to non‐transfer identification and fast iterative sample transfer identification effectively reduces the calculation amount when dealing with low quality measurement data.

Effect of Age and Gender on Categorical Perception of Vocal Emotion Under Tonal Language Background.

Categorical perception (CP) manifests in various aspects of human cognition. While there is mounting evidence for CP in facial emotions, CP in vocal emotions remains understudied. The current study attempted to test whether individuals with a tonal language background perceive vocal emotions categorically and to examine how factors such as gender and age influence the plasticity of these perceptual categories. This study examined the identification and discrimination performance of 24 Mandarin-speaking children (14 boys and 10 girls) and 32 adults (16 males and 16 females) when they were presented with three vocal emotion continua. Speech stimuli in each continuum consisted of 11 resynthesized Mandarin disyllabic words. CP phenomena were detected when Mandarin participants perceived vocal emotions. We further found the modulating effect of age and gender in vocal emotion categorization. Our results demonstrate for the first time that a categorical strategy is used by Mandarin speakers when perceiving vocal emotions. Furthermore, our findings reveal that the categorization ability of vocal emotions follows a prolonged course of development and the maturation patterns differ across genders. This study opens a promising line of research for investigating how sensory features are mapped to higher order perception and provides implications for our understanding of clinical populations characterized by altered emotional processing. https://doi.org/10.23641/asha.27204057.

A rate-them-all lineup procedure increases information but reduces discriminability.

Prior research has investigated ways to optimize identification performance, but an open question concerns exactly what variable should be optimized. One reasonable way to optimize performance is to maximize discriminability, which is achieved by increasing correct identifications of guilty suspects while simultaneously decreasing false identifications of innocent suspects. Another reasonable way to optimize performance is to maximize the information about the guilt or innocence of the suspect, which is best achieved by ensuring that a confidence rating is always made to the suspect. In a typical lineup, however, limited information about the suspect is obtained if the witness picks a filler or rejects the lineup. One proposed solution to that problem is to have the witness provide a confidence rating to every member of the lineup (a rate-them-all lineup). But what effect, if any, does a rate-them-all procedure have on discriminability? To answer that question, we compared a rate-them-all lineup procedure to standard simultaneous lineup and showup procedures using receiver operating characteristic analysis. In terms of discriminability, the rate-them-all procedure was diagnostically inferior to both. A reasonable goal for future research is to make use of theoretical models of eyewitness identification to simultaneously maximize both discriminability and information gain. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

Inertial Measurement Unit-Based Frozen Shoulder Identification from Daily Shoulder Tasks Using Machine Learning Approaches.

Frozen shoulder (FS) is a common shoulder condition accompanied by shoulder pain and a loss of shoulder range of motion (ROM). The typical clinical assessment tools such as questionnaires and ROM measurement are susceptible to subjectivity and individual bias. To provide an objective evaluation for clinical assessment, this study proposes an inertial measurement unit (IMU)-based identification system to automatically identify shoulder tasks whether performed by healthy subjects or FS patients. Two groups of features (time-domain statistical features and kinematic features), seven machine learning (ML) techniques, and two deep learning (DL) models are applied in the proposed identification system. For the experiments, 24 FS patients and 20 healthy subjects were recruited to perform five daily shoulder tasks with two IMUs attached to the arm and the wrist. The results demonstrate that the proposed system using deep learning presented the best identification performance using all features. The convolutional neural network achieved the best identification accuracy of 88.26%, and the multilayer perceptron obtained the best F1 score of 89.23%. Further analysis revealed that the identification performance based on wrist features had a higher accuracy compared to that based on arm features. The system's performance using time-domain statistical features has better discriminability in terms of identifying FS compared to using kinematic features. We demonstrate that the implementation of the IMU-based identification system using ML is feasible for FS assessment in clinical practice.

Long-Range Bird Species Identification Using Directional Microphones and CNNs

This study explores the integration of directional microphones with convolutional neural networks (CNNs) for long-range bird species identification. By employing directional microphones, we aimed to capture high-resolution audio from specific directions, potentially improving the clarity of bird calls over extended distances. Our approach involved processing these recordings with CNNs trained on a diverse dataset of bird calls. The results demonstrated that the system is capable of systematically identifying bird species up to 150 m, reaching 280 m for species vocalizing at frequencies greater than 1000 Hz and clearly distinct from background noise. The furthest successful detection was obtained at 510 m. While the method showed promise in enhancing the identification process compared to traditional techniques, there were notable limitations in the clarity of the audio recordings. These findings suggest that while the integration of directional microphones and CNNs for long-range bird species identification is promising, further refinement is needed to fully realize the benefits of this approach. Future efforts should focus on improving the audio-capture technology to reduce ambient noise and enhance the system’s overall performance in long-range bird species identification.

LTFM: Long-tail few-shot module with loose coupling strategy for mineral spectral identification

In recent years, deep learning methods have exhibited superior performance in mineral identification when especially compared with conventional machine learning methods such as Support Vector Machine (SVM) and Partial Least Squares (PLS). Nevertheless, almost all of these deep learning methods pay more attention to improving and designing network structures, while neglecting the phenomenon of long-tail distribution in spectral data due to the inconsistency of ore distribution and the scarcity of several natural minerals. To alleviate the interference of majority categories on minority categories, we propose Long-Tail Few-shot Module (LTFM) which is inspired by rethinking the fashionable decoupling strategy that conducts primary representation learning and further classifier retrained on mineral spectral data. In particular, LTFM serves as a multi-expert mode, where these experts are respectively specialized in improving feature representation learning, mitigating the long-tail effect, and alleviating the interference of few shots. Additionally, the loose coupling learning strategy is introduced to facilitate primary representation learning and the subsequent additional experts to inherit this knowledge. Experiments on two publicly available spectral datasets show that the proposed LTFM significantly outperforms existing methods. In the end, extensive ablation studies are conducted to investigate the effectiveness, correctness, and robustness of our proposal.

Role transition, investment practice and risk Management of Peer-to-peer Lending Investors: Based on the perspective of investor learning

Online peer-to-peer (P2P) lending has emerged as a result of the integration of finance and technology. Within the P2P lending market, individual investors have the opportunity to make repeated investments and engage in two-way transactions involving loans and investments. This setting provides an ideal scenario for studying the impact of different learning types on investors' risk management. Drawing on transaction data from the online lending platform Renrendai.com, we empirically examine the influence of role transition and investment practice on investors' risk identification, risk preference, and investment performance. We introduce variables related to loans and investment portfolios to analyze the effects. Our findings reveal that investment practice has a greater effect than role transition in reducing the investment default of online lending investors. Additionally, role transition is associated with a decrease in the investment amount set by investors and a reduced reliance on borrowers' credit ratings. Historical investment experience also leads to a decrease in the investment amount, while experiences of past investment failures improve investors' reference to borrowers' credit ratings. Furthermore, we observe that borrowing experience and previous investment failures contribute to a lower probability and proportion of high-risk loans in the portfolio. We shed light on the “black box” of investor learning and expand the research methodology for understanding investor behavior characteristics in the Fintech era.

YOLOv5s-Based Image Identification of Stripe Rust and Leaf Rust on Wheat at Different Growth Stages.

Stripe rust caused by Puccinia striiformis f. sp. tritici and leaf rust caused by Puccinia triticina, are two devastating diseases on wheat, which seriously affect the production safety of wheat. Timely detection and identification of the two diseases are essential for taking effective disease management measures to reduce wheat yield losses. To realize the accurate identification of wheat stripe rust and wheat leaf rust during the different growth stages, in this study, the image-based identification of wheat stripe rust and wheat leaf rust during different growth stages was investigated based on deep learning using image processing technology. Based on the YOLOv5s model, we built identification models of wheat stripe rust and wheat leaf rust during the seedling stage, stem elongation stage, booting stage, inflorescence emergence stage, anthesis stage, milk development stage, and all the growth stages. The models were tested on the different testing sets in the different individual growth stages and in all the growth stages. The results showed that the models performed differently in disease image identification. The model based on the disease images acquired during an individual growth stage was not suitable for the identification of the disease images acquired during the other individual growth stages, except for the model based on the disease images acquired during the milk development stage, which had acceptable identification performance on the testing sets in the anthesis stage and the milk development stage. In addition, the results demonstrated that wheat growth stages had a great influence on the image identification of the two diseases. The model built based on the disease images acquired in all the growth stages produced acceptable identification results. Mean F1 Score values between 64.06% and 79.98% and mean average precision (mAP) values between 66.55% and 82.80% were achieved on each testing set composed of the disease images acquired during an individual growth stage and on the testing set composed of the disease images acquired during all the growth stages. This study provides a basis for the image-based identification of wheat stripe rust and wheat leaf rust during the different growth stages, and it provides a reference for the accurate identification of other plant diseases.

Single-neuron representations of odours in the human brain

Olfaction is a fundamental sensory modality that guides animal and human behaviour1,2. However, the underlying neural processes of human olfaction are still poorly understood at the fundamental—that is, the single-neuron—level. Here we report recordings of single-neuron activity in the piriform cortex and medial temporal lobe in awake humans performing an odour rating and identification task. We identified odour-modulated neurons within the piriform cortex, amygdala, entorhinal cortex and hippocampus. In each of these regions, neuronal firing accurately encodes odour identity. Notably, repeated odour presentations reduce response firing rates, demonstrating central repetition suppression and habituation. Different medial temporal lobe regions have distinct roles in odour processing, with amygdala neurons encoding subjective odour valence, and hippocampal neurons predicting behavioural odour identification performance. Whereas piriform neurons preferably encode chemical odour identity, hippocampal activity reflects subjective odour perception. Critically, we identify that piriform cortex neurons reliably encode odour-related images, supporting a multimodal role of the human piriform cortex. We also observe marked cross-modal coding of both odours and images, especially in the amygdala and piriform cortex. Moreover, we identify neurons that respond to semantically coherent odour and image information, demonstrating conceptual coding schemes in olfaction. Our results bridge the long-standing gap between animal models and non-invasive human studies and advance our understanding of odour processing in the human brain by identifying neuronal odour-coding principles, regional functional differences and cross-modal integration.

Implicit memory reduced selectively for negative words with aging.

Disproportionally better memory for positive versus negative information (mnemonic positivity effect, MPE) in older versus younger adults has been reported on tests of explicit memory (direct, intentional) as measured by recall and recognition. The purpose of this investigation was to examine whether the MPE would be observed for implicit memory (indirect, unintentional) under conditions where, based on previous research using single words, it was expected that the MPE for explicit memory would be absent. This study investigated the influence of age on explicit and implicit memory for positive, negative, and neutral single words as measured by yes/no recognition and word identification on 24 older adults (aged 66-85) and 24 younger adults (aged 18-37) recruited from community centers in South Boston, Massachusetts. Older adults had lower recognition memory accuracy for positive, negative, and neutral words than younger adults, and, consistent with most prior studies, did not exhibit an explicit memory MPE for single words. For both groups, recognition accuracy was greatest for negative words, and was similar for positive and neutral words. In contrast, older adults exhibited implicit repetition priming, as measured by superior identification performance for repeated words, that was similar to younger adults for positive and neutral words. In younger adults, implicit memory was significantly greater for negative words than for positive and neutral words, whereas in older adults there were no significant differences in implicit memory for negative, positive, and neutral words. Therefore, selectively reduced priming for negative words in older adults was found in the context of enhanced priming for negative words in the younger adults. These findings show that there was an implicit memory MPE in older adults for words even under conditions where there was no explicit memory MPE in the older adults. Dampening of negative valence implicit memory with aging expands the perimeter of the age-related positivity framework.

DeepCOVIDNet-CXR: deep learning strategies for identifying COVID-19 on enhanced chest X-rays.

COVID-19 is one of the recent major epidemics, which accelerates its mortality and prevalence worldwide. Most literature on chest X-ray-based COVID-19 analysis has focused on multi-case classification (COVID-19, pneumonia, and normal) by the advantages of Deep Learning. However, the limited number of chest X-rays with COVID-19 is a prominent deficiency for clinical relevance. This study aims at evaluating COVID-19 identification performances using adaptive histogram equalization (AHE) to feed the ConvNet architectures with reliable lung anatomy of airways. We experimented with balanced small- and large-scale COVID-19 databases using left lung, right lung, and complete chest X-rays with various AHE parameters. On multiple strategies, we applied transfer learning on four ConvNet architectures (MobileNet, DarkNet19, VGG16, and AlexNet). Whereas DarkNet19 reached the highest multi-case identification performance with an accuracy rate of 98.26 % on the small-scale dataset, VGG16 achieved the best generalization performance with an accuracy rate of 95.04 % on the large-scale dataset. Our study is one of the pioneering approaches that analyses 3615 COVID-19 cases and specifies the most responsible AHE parameters for ConvNet architectures in the multi-case classification.

High-dimensional Biomarker Identification for Scalable and Interpretable Disease Prediction via Machine Learning Models.

Omics data generated from high-throughput technologies and clinical features jointly impact many complex human diseases. Identifying key biomarkers and clinical risk factors is essential for understanding disease mechanisms and advancing early disease diagnosis and precision medicine. However, the high-dimensionality and intricate associations between disease outcomes and omics profiles present significant analytical challenges. To address these, we propose an ensemble data-driven biomarker identification tool, Hybrid Feature Screening (HFS), to construct a candidate feature set for downstream advanced machine learning models. The pre-screened candidate features from HFS are further refined using a computationally efficient permutation-based feature importance test, forming the comprehensive High-dimensional Feature Importance Test (HiFIT) framework. Through extensive numerical simulations and real-world applications, we demonstrate HiFIT's superior performance in both outcome prediction and feature importance identification. An R package implementing HiFIT is available on GitHub ( https://github.com/BZou-lab/HiFIT ).

An exploratory study on automatic identification of assumptions in the development of deep learning frameworks

ContextStakeholders constantly make assumptions in the development of deep learning (DL) frameworks. These assumptions are related to various types of software artifacts (e.g., requirements, design decisions, and technical debt) and can turn out to be invalid, leading to system failures. Existing approaches and tools for assumption management usually depend on manual identification of assumptions. However, assumptions are scattered in various sources (e.g., code comments, commits, pull requests, and issues) of DL framework development, and manually identifying assumptions has high costs (e.g., time and resources). ObjectiveThe objective of the study is to evaluate different classification models for the purpose of identification with respect to assumptions from the point of view of developers and users in the context of DL framework projects (i.e., issues, pull requests, and commits) on GitHub. MethodFirst, we constructed a new and largest dataset (i.e., the AssuEval dataset) of assumptions collected from the TensorFlow and Keras repositories on GitHub. Then we explored the performance of seven non-transformers based models (e.g., Support Vector Machine, Classification and Regression Trees), the ALBERT model, and three decoder-only models (i.e., ChatGPT, Claude, and Gemini) for identifying assumptions on the AssuEval dataset. ResultsThe study results show that ALBERT achieves the best performance (f1-score: 0.9584) for identifying assumptions on the AssuEval dataset, which is much better than the other models (the 2nd best f1-score is 0.8858, achieved by the Claude 3.5 Sonnet model). Though ChatGPT, Claude, and Gemini are popular models, we do not recommend using them to identify assumptions in DL framework development because of their low performance. Fine-tuning ChatGPT, Claude, Gemini, or other language models (e.g., Llama3, Falcon, and BLOOM) specifically for assumptions might improve their performance for assumption identification. ConclusionsThis study provides researchers with the largest dataset of assumptions for further research (e.g., assumption classification, evaluation, and reasoning) and helps researchers and practitioners better understand assumptions and how to manage them in their projects (e.g., selection of classification models for identifying assumptions).

A novel method for identifying key nodes in multi-layer networks based on dynamic influence range and community importance

Identifying key nodes in multi-layer networks is a hot research topic in complex network science and has broad application prospects, such as in mining enterprises that significantly affecting multi-layer industrial chains. Unlike single-layer networks, nodes in multi-layer networks exhibit heterogeneity due to varying connections and locations. There are also correlations between different network layers, which is particularly evident in industrial chains where companies operate across multiple layers of production, supply and distribution. It is necessary to consider the impact of these layers on the global performance of key node identification. In addition, due to changes in connections, the community structure of each network layer should be different, reflecting the dynamic nature of industrial collaborations and partnerships. However, existing research lacks the model that addresses the above problems. Therefore, this paper proposes a key node identification method based on Dynamic Influence Range and Community Importance (DIRCI), using both local and global information of the multi-layer network simultaneously. DIRCI determines the importance of nodes through three centrality measures: dynamic influence range-based centrality, network layer centrality and community-based centrality. Dynamic influence range-based centrality models node heterogeneity by combining the influence range of nodes and their neighbors with lower computational costs. Network layer centrality captures the corresponding importance for different network layers. Community-based centrality comprehensively considers the importance of community, the importance of each node within the community and between different communities. Experimental results for nineteen multi-layer networks show that DIRCI achieves better performance of key node identification than the latest algorithms.

Radio frequency interference identification using dual cross-attention and multi-scale feature fusing

Radio astronomy plays a very important role in promoting scientific progress and unraveling the mysteries of the universe. However, radio telescopes are inevitably affected by radio frequency interference (RFI) when receiving radio signals, which leads to a reduction in data quality and has a serious impact on the formation of correct scientific conclusions. Therefore, it is essential to identify the RFI present in the observational data. In order to effectively identify RFI, improve the existing RFI identification methods that suffer from missed detections, and enhance the performance of RFI identification, this paper proposes a novel method that combines a dual cross-attention mechanism with multi-scale feature fusion. Experimental studies were conducted using the observational data from the 40-meter radio telescope at the Yunnan Astronomical Observatory of the Chinese Academy of Sciences. The proposed method achieved scores of 92.49%, 83.90%, and 87.99% in terms of precision, recall, and F1−score, respectively. It outperformed existing methods (U-Net, RFI-Net, R-Net6, RFI-GAN, EMSCA-UNet) in recall and F1−score, effectively reducing the occurrence of missed detections and improving the overall performance of radio frequency interference identification.