Learning Accuracy Research Articles

497 Diagnostic Accuracy of Deep Learning in Predicting Renal Cell Carcinoma: A Meta-Analysis

497 Diagnostic Accuracy of Deep Learning in Predicting Renal Cell Carcinoma: A Meta-Analysis

A novel structure learning method of Bayesian networks based on the neighboring complete node ordering search

Learning the optimal structure of a Bayesian network (BN) from observational data has received considerable research attention. In most structure learning methods based on scoring and searching, the discrete space of BNs is used, and the scale of the network grows exponentially with the number of nodes n, that is, O(n!2Cn2). Any existing scoring function in the BN space has multiple peaks. Thus, when using heuristic algorithms, local optimum problems may occur while searching in this space. As the search process involves only a small part of the space, the final learning result is not always ideal. In this study, the scoring and searching task is implemented in the complete node ordering space, and a novel neighbor operation is proposed for improving learning accuracy. This technique reduces the search space to O(n!) and facilitates optimized learning of a structure. Furthermore, techniques for efficiency optimization are proposed, which address the problems of high computational complexity and repetitive calculation of family scores.

Interpreting Mineral Deposit Genesis Classification with Decision Maps: A Case Study Using Pyrite Trace Elements

Abstract Machine learning improves geochemistry discriminant diagrams in classifying mineral deposit genetic types. However, the increasingly recognized ‘black box’ property of machine learning has been hampering the transparency of complex data analysis, leading to the challenge in deep geochemical interpretation. To address the issue, we revisited pyrite trace elements and propose to use ‘Decision Map’, a cutting-edge visualization technique for machine learning. This technique reveals mineral deposit classifications by visualizing the ‘decision boundaries’ of high-dimensional data, a concept crucial for model interpretation, active learning, and domain adaptation. In the context of geochemical data classification, it enables geologists to understand the relationship between geo-data and decision boundaries, assess prediction certainty, and observe the data distribution trends. This bridges the gap between the insightful properties of traditional discriminant diagrams and the high-dimensional efficiency of modern machine learning. Using pyrite trace element data, we construct a decision map for mineral deposit type classification, which maintains the accuracy of machine learning while adding valuable visualization insight. Additionally, we demonstrate two applications of decision maps. First, we show how decision maps can help resolve the genetic type dispute of a deposit whose data was not used in training the models. Second, we demonstrate how the decision maps can help understand the model, which further helps find indicator elements of pyrite. The recommended indicator elements by decision maps are consistent with geologists’ knowledge. This study confirms the decision map’s effectiveness in interpreting mineral genetic type classification problems. In geochemistry classification, it marks a shift from conventional machine learning to a visually insightful approach, thereby enhancing the geological understanding derived from the model. Furthermore, our work implies that decision maps could be applicable to diverse classification challenges in geosciences.

Impact of an Artificial Intelligence in Language Learning - A survey

Bully Scan, an artificial intelligence system for identifying offensive language on social media, is proposed in "A Natural Language Processing and Machine Learning-Based Framework to Automatically Identify Cyberbullying. This paradigm, which aims to reduce the negative impacts of cyberbullying and encourage healthy online interactions, is a critical step in using AI for social well-being. The paper, "Research and Practice of Hybrid Teaching Based on AI technology for Foreign Language Translation," offers a novel strategy for teaching foreign languages through the incorporation of AI. The project investigates a hybrid teaching approach that combines AI-powered language translation tools with conventional classroom training. This method seeks to improve accuracy and efficiency of language learning by providing real-time translation support. Through the use of AI technologies, such as machine learning and natural language processing, the system offers students helpful translation assistance, enhancing their educational experience. The study demonstrates encouraging outcomes in terms of raising students' proficiency and effectiveness in translation in a blended learning setting.
 The paper "Modular Design of English Pronunciation Level Evaluation System Based on Deep Learning Algorithm" offers a novel method for determining pronunciation levels in English by utilizing deep learning algorithms. The study uses techniques like support vector machines and BP neural networks to address the problem of computational intensity in language teaching technologies. Through the application of machine deep learning, the system seeks to improve the precision and efficacy of pronunciation level assessments, providing insightful information for the development of theories for foreign language instruction in the rapidly changing field of artificial intelligence. The study's modular design approach offers a viable foundation for enhancing pronunciation assessment in language instruction.

Bridging expertise with machine learning and automated machine learning in clinical medicine

In this issue of the Annals, Thirunavukarasu et al.’s systematic review on the clinical performance of automated machine learning (autoML) highlights its extensive applicability across 22 clinical specialties, showcasing its potential to redefine healthcare by making artificial intelligence (AI) technologies accessible to those without advanced computational skills.1 This enables the development of effective AI models that could rival or exceed the accuracy of traditional machine learning (ML) approaches and human diagnostic methods.

RECURSIVE NEURAL NETWORK AS A HIGH-SPEED PLATE COLLISION EMULATOR

Based on a database obtained using a high-speed plate impact model that relates impact parameters and material model parameters to the free surface velocity profile, the study compares the learning process and accuracy of a feedforward artificial neural network and a recursive neural network. A recursive neural network provides a significantly greater accuracy and requires less training time. Using a recursive neural network as a fast model emulator and Bayesian calibration can make it possible to solve the inverse problem of determining the substance model parameters from the free surface velocity profile with a greater accuracy.

Accuracy of machine learning in the diagnosis of odontogenic cysts and tumors: a systematic review and meta-analysis.

The recent impact of artificial intelligence in diagnostic services has been enormous. Machine learning tools offer an innovative alternative to diagnose cysts and tumors radiographically that pose certain challenges due to the near similar presentation, anatomical variations, and superimposition. It is crucial that the performance of these models is evaluated for their clinical applicability in diagnosing cysts and tumors. A comprehensive literature search was carried out on eminent databases for published studies between January 2015 and December 2022. Studies utilizing machine learning models in the diagnosis of odontogenic cysts or tumors using Orthopantomograms (OPG) or Cone Beam Computed Tomographic images (CBCT) were included. QUADAS-2 tool was used for the assessment of the risk of bias and applicability concerns. Meta-analysis was performed for studies reporting sufficient performance metrics, separately for OPG and CBCT. 16 studies were included for qualitative synthesis including a total of 10,872 odontogenic cysts and tumors. The sensitivity and specificity of machine learning in diagnosing cysts and tumors through OPG were 0.83 (95% CI 0.81-0.85) and 0.82 (95% CI 0.81-0.83) respectively. Studies utilizing CBCT noted a sensitivity of 0.88 (95% CI 0.87-0.88) and specificity of 0.88 (95% CI 0.87-0.89). Highest classification accuracy was 100%, noted for Support Vector Machine classifier. The results from the present review favoured machine learning models to be used as a clinical adjunct in the radiographic diagnosis of odontogenic cysts and tumors, provided they undergo robust training with a huge dataset. However, the arduous process, investment, and certain ethical concerns associated with the total dependence on technology must be taken into account. Standardized reporting of outcomes for diagnostic studies utilizing machine learning methods is recommended to ensure homogeneity in assessment criteria, facilitate comparison between different studies, and promote transparency in research findings.

Game-Theoretic Unlearnable Example Generator

Unlearnable example attacks are data poisoning attacks aiming to degrade the clean test accuracy of deep learning by adding imperceptible perturbations to the training samples, which can be formulated as a bi-level optimization problem. However, directly solving this optimization problem is intractable for deep neural networks. In this paper, we investigate unlearnable example attacks from a game-theoretic perspective, by formulating the attack as a nonzero sum Stackelberg game. First, the existence of game equilibria is proved under the normal setting and the adversarial training setting. It is shown that the game equilibrium gives the most powerful poison attack in that the victim has the lowest test accuracy among all networks within the same hypothesis space when certain loss functions are used. Second, we propose a novel attack method, called the Game Unlearnable Example (GUE), which has three main gradients. (1) The poisons are obtained by directly solving the equilibrium of the Stackelberg game with a first-order algorithm. (2) We employ an autoencoder-like generative network model as the poison attacker. (3) A novel payoff function is introduced to evaluate the performance of the poison. Comprehensive experiments demonstrate that GUE can effectively poison the model in various scenarios. Furthermore, the GUE still works by using a relatively small percentage of the training data to train the generator, and the poison generator can generalize to unseen data well. Our implementation code can be found at https://github.com/hong-xian/gue.

Label enhancement via manifold approximation and projection with graph convolutional network

Label enhancement (LE) aims to enrich logical labels into their corresponding label distributions. But existing LE algorithms fail to fully leverage the structural information in the feature space to improve LE learning. To address this key issue, we first apply manifold learning to map the relatedness between low-dimensional feature samples to the label space. Based on the smoothness assumption of manifolds, the implicit correlation between low-dimensional feature and label spaces effectively promotes the LE process, enabling the learning model to accurately capture the mapping relationship between feature and label manifolds. This leads to an LE based on feature representation (LEFR) algorithm. We also propose an LE algorithm based on graph convolutional network (GCN), called LE-GCN. Inspired by the relationship between threshold connections and label connections, we extend GCN to the LE field for the first time to fully exploit the hidden relationships between nodes and labels. By enhancing node information with threshold connections and label connections, the label learning accuracy reaches a new level. Experiments on real-world datasets show that our LEFR and LE-GCN outperform several state-of-the-art LE algorithms.

FedCSL: A Scalable and Accurate Approach to Federated Causal Structure Learning

As an emerging research direction, federated causal structure learning (CSL) aims at learning causal relationships from decentralized data across multiple clients while preserving data privacy. Existing federated CSL algorithms suffer from scalability and accuracy issues, since they require computationally expensive CSL algorithms to be executed at each client. Furthermore, in real-world scenarios, the number of samples held by each client varies significantly, and existing methods still assign equal weights to the learned structural information from each client, which severely harms the learning accuracy of those methods. To address these two limitations, we propose FedCSL, a scalable and accurate method for federated CSL. Specifically, FedCSL consists of two novel strategies: (1) a federated local-to-global learning strategy that enables FedCSL to scale to high-dimensional data for tackling the scalability issue, and (2) a novel weighted aggregation strategy that does not rely on any complex encryption techniques while preserving data privacy for tackling the accuracy issue. Extensive experiments on benchmark datasets, high-dimensional synthetic datasets and a real-world dataset verify the efficacy of the proposed FedCSL method. The source code is available at https://github.com/Xianjie-Guo/FedCSL.

An imbalanced small sample slab defect recognition method based on image generation

The surface quality of slabs has a profound impact on the subsequent performance and quality of steel. Accurate detection and classification of surface defects are crucial for improving steel quality. However, online visual inspection in the steel industry faces challenges such as sparse data, unclear features, sample imbalances, and impeding the application of deep learning in slab defect detection. To address these challenges, this paper introduces a novel approach utilizing the Slab Defect Generation (SDG) network to generate additional images by training on imbalanced small samples of slab defects, thereby constructing a comprehensive defect database. This methodology significantly enhances the efficiency and accuracy of deep learning in slab defect detection. Furthermore, an improved loss function is proposed specifically for the You Only Look Once version 5 (YOLOv5) model to augment the training of challenging samples. Experimental validation confirms the superiority of our method in terms of both image generation quality and defect-recognition compared to State-Of-The-Art (SOTA) models. The accuracy achieved by our approach is 98.6 %, outperforming the second-ranking model with an accuracy of 90.1 %.Index Terms—Data Enhancement, Deep Learning, Defect Recognition, Machine Vision, Neural Networks, Small sample learning.

Monitoring and Early Warning Method of Debris Flow Expansion Behavior Based on Improved Genetic Algorithm and Bayesian Network

Based on an improved genetic algorithm and debris flow disaster monitoring network, this study examines the monitoring and early warning method of debris flow expansion behavior, divides the risk of debris flow disaster, and provides a scientific basis for emergency rescue and post-disaster recovery. The function of the debris flow disaster monitoring network of the spreading behavior disaster chain is constructed. According to the causal reasoning of debris flow disaster monitoring information, the influence factors of debris flow, such as rainfall intensity and duration, are selected as the inputs of the Bayesian network, and the probability of a debris flow disaster is obtained. The probability is compared with the historical data threshold to complete the monitoring and early warning of debris flow spreading behavior. Innovatively, by introducing niche technology to improve traditional genetic algorithms by learning Bayesian networks, the optimization efficiency and convergence speed of genetic algorithms are improved, and the robustness of debris flow monitoring and warning is enhanced. The experimental results show that this method divides debris flow disasters into the following five categories based on their danger: low-risk area, medium-risk area, high-risk area, higher-risk area, and Very high-risk area. It accurately monitors the expansion of debris flows and completes early warning. The disaster management department can develop emergency rescue and post-disaster recovery strategies based on early warning results, thus providing a scientific basis for debris flow disasters. The improved genetic algorithm has a higher learning efficiency, a higher accuracy, a faster convergence speed, and higher advantages in learning time and accuracy of the Bayesian network structure.

BGOA-TVG: Binary Grasshopper Optimization Algorithm with Time-Varying Gaussian Transfer Functions for Feature Selection.

Feature selection aims to select crucial features to improve classification accuracy in machine learning and data mining. In this paper, a new binary grasshopper optimization algorithm using time-varying Gaussian transfer functions (BGOA-TVG) is proposed for feature selection. Compared with the traditional S-shaped and V-shaped transfer functions, the proposed Gaussian time-varying transfer functions have the characteristics of a fast convergence speed and a strong global search capability to convert a continuous search space to a binary one. The BGOA-TVG is tested and compared to S-shaped and V-shaped binary grasshopper optimization algorithms and five state-of-the-art swarm intelligence algorithms for feature selection. The experimental results show that the BGOA-TVG has better performance in UCI, DEAP, and EPILEPSY datasets for feature selection.

Grouped machine learning methods for predicting rock mass parameters in a tunnel boring machine‐driven tunnel based on fuzzy C‐means clustering

AbstractTo guarantee safe and efficient tunneling of a tunnel boring machine (TBM), rapid and accurate judgment of the rock mass condition is essential. Based on fuzzy C‐means clustering, this paper proposes a grouped machine learning method for predicting rock mass parameters. An elaborate data set on field rock mass is collected, which also matches field TBM tunneling. Meanwhile, target stratum samples are divided into several clusters by fuzzy C‐means clustering, and multiple submodels are trained by samples in different clusters with the input of pretreated TBM tunneling data and the output of rock mass parameter data. Each testing sample or newly encountered tunneling condition can be predicted by multiple submodels with the weight of the membership degree of the sample to each cluster. The proposed method has been realized by 100 training samples and verified by 30 testing samples collected from the C1 part of the Pearl Delta water resources allocation project. The average percentage error of uniaxial compressive strength and joint frequency (Jf) of the 30 testing samples predicted by the pure back propagation (BP) neural network is 13.62% and 12.38%, while that predicted by the BP neural network combined with fuzzy C‐means is 7.66% and 6.40%, respectively. In addition, by combining fuzzy C‐means clustering, the prediction accuracies of support vector regression and random forest are also improved to different degrees, which demonstrates that fuzzy C‐means clustering is helpful for improving the prediction accuracy of machine learning and thus has good applicability. Accordingly, the proposed method is valuable for predicting rock mass parameters during TBM tunneling.

FFYOLO: A Lightweight Forest Fire Detection Model Based on YOLOv8

Forest is an important resource for human survival, and forest fires are a serious threat to forest protection. Therefore, the early detection of fire and smoke is particularly important. Based on the manually set feature extraction method, the detection accuracy of the machine learning forest fire detection method is limited, and it is unable to deal with complex scenes. Meanwhile, most deep learning methods are difficult to deploy due to high computational costs. To address these issues, this paper proposes a lightweight forest fire detection model based on YOLOv8 (FFYOLO). Firstly, in order to better extract the features of fire and smoke, a channel prior dilatation attention module (CPDA) is proposed. Secondly, the mixed-classification detection head (MCDH), a new detection head, is designed. Furthermore, MPDIoU is introduced to enhance the regression and classification accuracy of the model. Then, in the Neck section, a lightweight GSConv module is applied to reduce parameters while maintaining model accuracy. Finally, the knowledge distillation strategy is used during training stage to enhance the generalization ability of the model and reduce the false detection. Experimental outcomes demonstrate that, in comparison to the original model, FFYOLO realizes an mAP0.5 of 88.8% on a custom forest fire dataset, which is 3.4% better than the original model, with 25.3% lower parameters and 9.3% higher frames per second (FPS).

An enhanced automated machine learning model for optimizing cycle-to-cycle variation in hydrogen-enriched methanol engines

Renewable methanol and hydrogen have emerged as great potential alternative fuels for internal combustion engines in recent research. Hydrogen exhibits the ability to mitigate cycle-to-cycle variations in methanol engines thanks to its rapid combustion rate. Furthermore, the integration of machine learning with genetic algorithms provides a solution to achieve efficient and clean combustion in hydrogen-enriched methanol engines. Nevertheless, this approach is constrained due to the limited predictive accuracy in small-sample learning. This study aims at introducing an enhanced model based on the automated machine learning framework, and provide a comparative analysis of the enhanced model and computational fluid dynamics simulations in terms of multi-objective optimization performance and computational time. Additionally, this study further considered the effect of engine cycle-to-cycle variations on performance, ultimately attaining optimal operating parameters for stable combustion. The results indicate that datasets actively created by genetic algorithms are more effective than those generated through Latin hypercube sampling. The introduction of a classification model for the identification of misfires has enhanced the efficiency of dataset utilization. Through feature construction, the automated machine learning model demonstrates improvements in R-square and mean-squared error for all parameters on the test dataset. Furthermore, by employing a larger overall population size, the automated machine learning model has not only achieved a higher-quality Pareto fronts compared to the computational fluid dynamics method but has also significantly reduced computation time, particularly upon introducing additional engine cycle-to-cycle variations constraints.

Prediction of liquefaction of gravelly soils based on a cost-sensitive Bayesian network combined with rough set weighting

Existing liquefaction prediction approaches for gravelly soils rely on in situ test data, which neglects the in-suit test cost sensitivity. Furthermore, these approaches do not consider the effect of the weights of the factors on the model performance. Therefore, in this paper, a cost-sensitive Bayesian network with rough set theory (CSW-BN) and model hierarchies and factor weighting are combined to address the two aforementioned challenges in research on gravelly soil liquefaction prediction. A two-layer model is constructed using the proposed CSW-BN approach. The first layer of the model reduces the data costs by utilizing conventional soil parameters, without necessitating specific in situ tests such as shear wave velocity tests. In the second layer of the model, variable weights calculated using rough set theory are incorporated into the parameter learning to enhance the generalization of the model. The results show that the learning and prediction accuracies of the new approach are 0.968 and 0.95, respectively. Compared with existing models, the proposed method not only reduces the number of required in situ tests but also improves the prediction accuracy and interpretability. Furthermore, the effectiveness of the CSW-BN model is evaluated using new cases for the 2008 Wenchuan earthquake, with a prediction accuracy of 0.89. Finally, a software tool is developed using Visual Basic for Applications programming to facilitate its application in engineering practice.

Research on road Damage Detection Based on Improved YOLOv8

Aiming at the low accuracy of deep learning in road damage detection, a road damage detection method based on improved YOLOv8 is proposed. Firstly, SENetv2 attention module was added to the backbone network of YOLOv8 to improve the model's feature learning ability. Secondly, in the neck network, GSConv is used to replace the common convolutional module to reduce the complexity of the model and improve the accuracy. Finally, the loss function is changed to Wise-iou to reduce the impact of detection due to a small number of low-quality instances. The experimental results show that compared with the traditional YOLOv8, the mAP50 of this model is increased by 2.1 percentage points, and the detection effect is good, which can meet the requirements of accurate detection.

Unveiling the Impact of Extreme Learning Machine in the Defence and Military Sector

Among the most well-known machine learning algorithms, Extreme Learning Machine (ELM) has seen widespread use across a variety of fields, including the defence and military industries. For problems like sluggish technique and iteratively altering the hidden layer’s parameters to optimise the efficiency of the gradient descent approach, a cutting-edge machine learning algorithm known as the ELM has been developed. Depending on the specific objective and circumstance, the Extreme Learning Machine (ELM) may be more appropriate than Deep Neural Network (DNN) techniques. The models constructed in this manner perform quite well in generalisation. The following three goals are emphasised in its unconventional structure: 1) A great degree of accuracy in learning 2) less need for direct human involvement 3) an extremely rapid rate of learning, and moreover, it provides an optimal response for the whole world. As a result of its quick training, flexibility, and resilience, the Extreme Learning Machine (ELM) has several uses in this field, including target detection and tracking, image and signal processing, cybersecurity and intrusion detection, decision support systems, pattern recognition and classification, etc. According to our findings, the ELM approach was used with low training time and the testing accuracy is excellent. Also, this study presents the contribution of the revolutionary machine learning algorithm ELM to the defence and military sectors.

Accuracy of machine learning in the preoperative identification of ovarian borderline tumors: a meta-analysis

The objective of this study is to explore the diagnostic value of machine learning (ML) in borderline ovarian tumors through meta-analysis. Pubmed, Embase, Web of Science, and Cochrane Library databases were comprehensively retrieved from database inception untill February 16, 2023. The Prediction Model Risk of Bias Assessment Tool (PROBAST) was adopted to evaluate the risk of bias in the original studies. Sub-group analyses of ML were conducted according to clinical features and radiomics features. We separately discussed the discriminative value of ML for borderline vs benign and borderline vs malignant tumors. Eighteen studies involving 12,778 subjects were included in our analysis. The modeling variables mainly consisted of radiomics features (n=13) and a small number of clinical features (n=5). When distinguishing between borderline and benign tumors, the ML model based on radiomic features achieved a c-index of 0.782 (95% CI: 0.732-0.831), sensitivity of 0.75 (95% CI: 0.67-0.82), and specificity of 0.75 (95% CI: 0.67-0.81) in the validation set. When distinguishing between borderline and malignant tumors, the ML model based on radiomic features achieved a c-index of 0.916 (95% CI: 0.891-0.940), sensitivity of 0.86 (95% CI: 0.78-0.91), and specificity of 0.88 (95% CI: 0.82-0.92) in the validation set. In addition, we analyzed the discriminatory ability of radiologists and found that their sensitivity was 0.26 (95% CI: 0.12-0.46) and specificity was 0.94 (95% CI: 0.90-0.97). ML has tremendous potential in the preoperative diagnosis and differentiation of borderline ovarian tumors and may be more accurate than radiologists in diagnosing and differentiating borderline ovarian tumors.