Search Results (12)

Glaucoma is an irreversible neurodegenerative disease that affects the optic nerve, leading to partial or complete vision loss. Early and accurate detection is crucial to prevent vision impairment, which necessitates the development of highly precise diagnostic tools. Deep learning (DL) has emerged as a promising approach for glaucoma diagnosis, where the model is trained on datasets of fundus images. To improve the detection accuracy, we propose a hybrid model for glaucoma detection that combines multiple DL models with two fine-tuning strategies and uses a majority voting scheme to determine the final prediction. In experiments, the hybrid model achieved a detection accuracy of 96.55%, a sensitivity of 98.84%, and a specificity of 94.32%. Integrating datasets was found to improve the performance compared to using them separately even with transfer learning. When compared to individual DL models, the hybrid model achieved a 20.69% improvement in accuracy compared to the best model when applied to a single dataset, a 13.22% improvement when applied with transfer learning across all datasets, and a 1.72% improvement when applied to all datasets. These results demonstrate the potential of hybrid DL models to detect glaucoma more accurately than individual models. Full article

Background: Digital breast tomosynthesis (DBT) is a valuable imaging modality for breast cancer detection; however, its interpretation remains time-consuming and subject to inter-reader variability. This study aimed to develop and evaluate two deep learning (DL) models based on transfer learning for the binary classification of breast lesions (benign vs. malignant) using DBT images to support clinical decision-making and risk stratification. Methods: In this retrospective monocentric study, 184 patients with histologically or clinically confirmed benign (107 cases, 41.8%) or malignant (77 cases, 58.2%) breast lesions were included. Each case underwent DBT with a single lesion manually segmented for radiomic analysis. Two convolutional neural network (CNN) architectures—ResNet50 and DenseNet201—were trained using transfer learning from ImageNet weights. A 10-fold cross-validation strategy with ensemble voting was applied. Model performance was evaluated through ROC–AUC, accuracy, sensitivity, specificity, PPV, and NPV. Results: The ResNet50 model outperformed DenseNet201 across most metrics. On the internal testing set, ResNet50 achieved a ROC–AUC of 63%, accuracy of 60%, sensitivity of 39%, and specificity of 75%. The DenseNet201 model yielded a lower ROC–AUC of 55%, accuracy of 55%, and sensitivity of 24%. Both models demonstrated relatively high specificity, indicating potential utility in ruling out malignancy, though sensitivity remained suboptimal. Conclusions: This study demonstrates the feasibility of using transfer learning-based DL models for lesion classification on DBT. While the overall performance was moderate, the results highlight both the potential and current limitations of AI in breast imaging. Further studies and approaches are warranted to enhance model robustness and clinical applicability. Full article

Accurately predicting the state of surface water quality is crucial for ensuring the sustainable use of water resources and environmental protection. This often requires a focus on the range of factors affecting water quality, such as physical and chemical parameters. Tree models, with their flexible tree-like structure and strong capability for partitioning and selecting influential water quality features, offer clear decision-making rules, making them suitable for this task. However, an individual decision tree model has limitations and cannot fully capture the complex relationships between all influencing parameters and water quality. Therefore, this study proposes a method combining ensemble tree models with voting algorithms to predict water quality classification. This study was conducted using five surface water monitoring sites in Qingdao, representing a portion of many municipal water environment monitoring stations in China, employing a single-factor determination method with stringent surface water standards. The soft voting algorithm achieved the highest accuracy of 99.91%, and the model addressed the imbalance in original water quality categories, reaching a Matthews Correlation Coefficient (MCC) of 99.88%. In contrast, conventional machine learning algorithms, such as logistic regression and K-nearest neighbors, achieved lower accuracies of 75.90% and 91.33%, respectively. Additionally, the model’s supervision of misclassified data demonstrated its good learning of water quality determination rules. The trained model was also transferred directly to predict water quality at 13 monitoring stations in Beijing, where it performed robustly, achieving an ensemble hard voting accuracy of 97.73% and an MCC of 96.81%. In many countries’ water environment systems, different water qualities correspond to different uses, and the magnitude of influencing parameters is directly related to water quality categories; critical parameters can even directly determine the quality category. Tree models are highly capable of handling nonlinear relationships and selecting important water quality features, allowing them to identify and exploit interactions between water quality parameters, which is especially important when multiple parameters together determine the water quality category. Therefore, there is significant motivation to develop tree model-based water quality prediction models. Full article

The abundance of powered semiconductor devices has increased with the introduction of renewable energy sources into the grid, causing power quality disturbances (PQDs). This represents a huge challenge for grid reliability and smart city infrastructures. Accurate detection and classification are important for grid reliability and consumers’ appliances in a smart city environment. Conventionally, power quality monitoring relies on trivial machine learning classifiers or signal processing methods. However, recent advancements have introduced Deep Convolution Neural Networks (DCNNs) as promising methods for the detection and classification of PQDs. These techniques have the potential to demonstrate high classification accuracy, making them a more appropriate choice for real-time operations in a smart city framework. This paper presents a voting ensemble approach to classify sixteen PQDs, using the DCNN architecture through transfer learning. In this process, continuous wavelet transform (CWT) is employed to convert one-dimensional (1-D) PQD signals into time–frequency images. Four pre-trained DCNN architectures, i.e., Residual Network-50 (ResNet-50), Visual Geometry Group-16 (VGG-16), AlexNet and SqeezeNet are trained and implemented in MATLAB, using images of four datasets, i.e., without noise, 20 dB noise, 30 dB noise and random noise. Additionally, we also tested the performance of ResNet-50 with a squeeze-and-excitation (SE) mechanism. It was observed that ResNet-50 with the SE mechanism has a better classification accuracy; however, it causes computational overheads. The classification performance is enhanced by using the voting ensemble model. The results indicate that the proposed scheme improved the accuracy (99.98%), precision (99.97%), recall (99.80%) and F1-score (99.85%). As an outcome of this work, it is demonstrated that ResNet-50 with the SE mechanism is a viable choice as a single classification model, while an ensemble approach further increases the generalized performance for PQD classification. Full article

Climate change, primarily driven by human activities such as burning fossil fuels, is causing significant long-term changes in temperature and weather patterns. To mitigate these impacts, there is an increased focus on renewable energy sources. However, optimizing power consumption through effective usage control and waste recycling also offers substantial potential for reducing energy demands. This study explores non-intrusive load monitoring (NILM) to estimate disaggregated energy consumption from a single household meter, leveraging advancements in deep learning such as convolutional neural networks. The study uses the UK-DALE dataset to extract and plot power consumption data from the main meter and identify five household appliances. Convolutional neural networks (CNNs) are trained with transfer learning using VGG16 and MobileNet. The models are validated, tested on split datasets, and combined using ensemble methods for improved performance. A new voting scheme for ensembles is proposed, named weighted average confidence voting (WeCV), and it is used to create combinations of the best 3 and 5 models and applied to NILM. The base models achieve up to 97% accuracy. The ensemble methods applying WeCV show an increased accuracy of 98%, surpassing previous state-of-the-art results. This study shows that CNNs with transfer learning effectively disaggregate household energy use, achieving high accuracy. Ensemble methods further improve performance, offering a promising approach for optimizing energy use and mitigating climate change. Full article

Machine learning models are used to identify crops in satellite data, which achieve high classification accuracy but do not necessarily have a high degree of transferability to new regions. This paper investigates the use of machine learning models for crop classification using Sentinel-2 imagery. It proposes a new testing methodology that systematically analyzes the quality of the spatial transfer of trained models. In this study, the classification results of Random Forest (RF), eXtreme Gradient Boosting (XGBoost), Stochastic Gradient Descent (SGD), Multilayer Perceptron (MLP), Support Vector Machines (SVM), and a Majority Voting of all models and their spatial transferability are assessed. The proposed testing methodology comprises 18 test scenarios to investigate phenological, temporal, spatial, and quantitative (quantitative regarding available training data) influences. Results show that the model accuracies tend to decrease with increasing time due to the differences in phenological phases in different regions, with a combined F1-score of 82% (XGBoost) when trained on a single day, 72% (XGBoost) when trained on the half-season, and 61% when trained over the entire growing season (Majority Voting). Full article

Deep learning chatbot research and development is exploding recently to offer customers in numerous industries personalized services. However, human resources are used to create a learning dataset for a deep learning chatbot. In order to augment this, the idea of neural question generation (NQG) has evolved, although it has restrictions on how questions can be expressed in different ways and has a finite capacity for question generation. In this paper, we propose an ensemble-type NQG model based on the text-to-text transfer transformer (T5). Through the proposed model, the number of generated questions for each single NQG model can be greatly increased by considering the mutual similarity and the quality of the questions using the soft-voting method. For the training of the soft-voting algorithm, the evaluation score and mutual similarity score weights based on the context and the question–answer (QA) dataset are used as the threshold weight. Performance comparison results with existing T5-based NQG models using the SQuAD 2.0 dataset demonstrate the effectiveness of the proposed method for QG. The implementation of the proposed ensemble model is anticipated to span diverse industrial fields, including interactive chatbots, robotic process automation (RPA), and Internet of Things (IoT) services in the future. Full article

Precision ventilation is developed to achieve thermal comfort for occupants in an office by creating micro-climate zones. The present study aims to achieve individual thermal comfort for occupants with different metabolic rates by using higher airspeeds for enhancing heat transfer. The variable jet interaction between two ACBs with JetCone adjustments cause higher velocity jets to reach different regions of the occupied zone. The colliding jets from the center of a thermal isolated room were moved towards different zones in an office configuration with a constant room temperature of 23 °C. This study was conducted for five different cases in a room divided into four zones according to occupants’ metabolic rates. The experimental and CFD results show that occupants facing symmetrical airflow distribution and with a constant 1.2 metabolic rate (Case 1) had a similar predicted mean vote (PMV) index. The zones with higher-metabolic-rate occupants, i.e., 1.4 met and 1.6 met in cases 2 and 3 were exposed to air velocities up to 0.4 and 0.5 m/s, respectively. In case 4, the air velocity in the single zone with 1.6 met occupants was raised to 0.6 m/s by targeted airflow distribution achieved by adjusting JetCones. These occupants with higher metabolic rates were kept thermally neutral, in the −0.5 to +0.5 PMV range, by pushing the high velocity colliding jets from the center towards them. In case 5, the results showed that precision ventilation can maintain the individual thermal comfort of up to three different zones (in the same office space) by exposing the occupants with metabolic rates of 1.2, 1.4, and 1.6 met to airspeeds of 0.15, 0.45, and 0.55 m/s, respectively. Full article

There are two main approaches to achieving proportional representation in elections: the single transferable vote and methods based on party lists. This paper discusses ways to use the single transferable vote while using some of the main features used with the largest remainder method, such as the electoral threshold. The investigation has shown that the Weighted Inclusive Gregory method is a suitable replacement for the largest remainder method when it is desirable to avoid wasted votes and to handle independent candidates in a straightforward way, but it is also desirable to keep the results as close to the ones achieved under the largest remainder method as possible. The investigation also led to the development of an algorithm for using the single transferable vote when preference lists are based on party lists, exploiting commonalities and symmetries between the patterns of preferences given in the votes. It has been shown that such an algorithm makes the calculations faster than the use of ordinary implementations of the single transferable vote when the numbers of seats and candidates are high, as commonly happens when methods based on party lists are used. Full article

Traditional methods for individual tree-crown (ITC) detection (image classification, segmentation, template matching, etc.) applied to very high-resolution remote sensing imagery have been shown to struggle in disparate landscape types or image resolutions due to scale problems and information complexity. Deep learning promised to overcome these shortcomings due to its superior performance and versatility, proven with reported detection rates of ~90%. However, such models still find their limits in transferability across study areas, because of different tree conditions (e.g., isolated trees vs. compact forests) and/or resolutions of the input data. This study introduces a highly replicable deep learning ensemble design for ITC detection and species classification based on the established single shot detector (SSD) model. The ensemble model design is based on varying the input data for the SSD models, coupled with a voting strategy for the output predictions. Very high-resolution unmanned aerial vehicles (UAV), aerial remote sensing imagery and elevation data are used in different combinations to test the performance of the ensemble models in three study sites with highly contrasting spatial patterns. The results show that ensemble models perform better than any single SSD model, regardless of the local tree conditions or image resolution. The detection performance and the accuracy rates improved by 3–18% with only as few as two participant single models, regardless of the study site. However, when more than two models were included, the performance of the ensemble models only improved slightly and even dropped. Full article

Technology scaling has led to an increase in density and capacity of on-chip caches. This has enabled higher throughput by enabling more low latency memory transfers. With the reduction in size of SRAMs and development of emerging technologies, e.g., STT-MRAM, for on-chip cache memories, reliability of such memories becomes a major concern. Traditional error correcting codes, e.g., Hamming codes and orthogonal Latin square codes, either suffer from high decoding latency, which leads to lower overall throughput, or high memory overhead. In this paper, a new single error correcting code based on a shared majority voting logic is presented. The proposed codes trade off decoding latency in order to improve the memory overhead posed by orthogonal Latin square codes. A latency optimization technique is also proposed which lowers the decoding latency by incurring a slight memory overhead. It is shown that the proposed codes achieve better redundancy compared to orthogonal Latin square codes. The proposed codes are also shown to achieve lower decoding latency compared to Hamming codes. Thus, the proposed codes achieve a balanced trade-off between memory overhead and decoding latency, which makes them highly suitable for on-chip cache memories which have stringent throughput and memory overhead constraints. Full article

Imaging spectroscopy is a promising tool for airborne tree species recognition in hyper-diverse tropical canopies. However, its widespread application is limited by the signal sensitivity to acquisition parameters, which may require new training data in every new area of application. This study explores how various pre-processing steps may improve species discrimination and species recognition under different operational settings. In the first experiment, a classifier was trained and applied on imaging spectroscopy data acquired on a single date, while in a second experiment, the classifier was trained on data from one date and applied to species identification on data from a different date. A radiative transfer model based on atmospheric compensation was applied with special focus on the automatic retrieval of aerosol amounts. The impact of spatial or spectral filtering and normalisation was explored as an alternative to atmospheric correction. A pixel-wise classification was performed with a linear discriminant analysis trained on individual tree crowns identified at the species level. Tree species were then identified at the crown scale based on a majority vote rule. Atmospheric corrections did not outperform simple statistical processing (i.e., filtering and normalisation) when training and testing sets were taken from the same flight date. However, atmospheric corrections became necessary for reliable species recognition when different dates were considered. Shadow masking improved species classification results in all cases. Single date classification rate was 83.9% for 1297 crowns of 20 tropical species. The loss of mean accuracy observed when using training data from one date to identify species at another date in the same area was limited to 10% when atmospheric correction was applied. Full article