Search

Over the past decades, drones have become more attainable by the public due to their widespread availability at affordable prices. Nevertheless, this situation sparks serious concerns in both the cyber and physical security domains, as drones can be employed for malicious activities with public safety threats. However, detecting drones instantly and efficiently is a very difficult task due to their tiny size and swift flights. This paper presents a novel drone detection method using deep convolutional learning and deep transfer learning. The proposed algorithm employs a new feature extraction network, which is added to the modified YOU ONLY LOOK ONCE version2 (YOLOv2) network. The feature extraction model uses bypass connections to learn features from the training sets and solves the "vanishing gradient" problem caused by the increasing depth of the network. The structure of YOLOv2 is modified by replacing the rectified linear unit (relu) with a leaky-relu activation function and adding an extra convolutional layer with a stride of 2 to improve the small object detection accuracy. Using leaky-relu solves the "dying relu" problem. The additional convolution layer with a stride of 2 reduces the spatial dimensions of the feature maps and helps the network to focus on larger contextual information while still preserving the ability to detect small objects. The model is trained with a custom dataset that contains various types of drones, airplanes, birds, and helicopters under various weather conditions. The proposed model demonstrates a notable performance, achieving an accuracy of 77% on the test images with only 5 million learnable parameters in contrast to the Darknet53 + YOLOv3 model, which exhibits a 54% accuracy on the same test set despite employing 62 million learnable parameters. [ABSTRACT FROM AUTHOR]

The diagnosis of different pathologies and stages of cancer using whole histopathology slide images (WSI) is the gold standard for determining the degree of tissue metastasis. The use of deep learning systems in the field of medical images, especially histopathology images, is becoming increasingly important. The training and optimization of deep neural network models involve fine-tuning parameters and hyperparameters such as learning rate, batch size (BS), and boost to improve the performance of the model in task-specific applications. Tuning hyperparameters is a major challenge in designing deep neural network models, having a large impact on the performance. This paper analyzes how the parameters and hyperparameters of a deep learning architecture affect the classification of colorectal cancer (CRC) histopathology images using the well-known VGG19 model. This paper also discusses the pre-processing of these images, such as the use of color normalization and stretching transformations on the data set. Among these hyperparameters, the most important neural network hyperparameter is the learning rate (LR). In this paper, different strategies for the optimization of LR are analyzed (both static and dynamic) and a new experiment based on the variation of LR is proposed (the relevance of dynamic strategies over fixed LR is highlighted), after each layer of the neural network together with decreasing variations according to the epochs. The results obtained are very remarkable, obtaining in the simulation an accurate system that achieves 96.4% accuracy on test images (for nine different tissue classes) using the triangular-cyclic learning rate. [ABSTRACT FROM AUTHOR]

The paper focuses on the hepatitis C virus (HCV) infection in Egypt, which has one of the highest rates of HCV in the world. The high prevalence is linked to several factors, including the use of injection drugs, poor sterilization practices in medical facilities, and low public awareness. This paper introduces a hyOPTGB model, which employs an optimized gradient boosting (GB) classifier to predict HCV disease in Egypt. The model's accuracy is enhanced by optimizing hyperparameters with the OPTUNA framework. Min-Max normalization is used as a preprocessing step for scaling the dataset values and using the forward selection (FS) wrapped method to identify essential features. The dataset used in the study contains 1385 instances and 29 features and is available at the UCI machine learning repository. The authors compare the performance of five machine learning models, including decision tree (DT), support vector machine (SVM), dummy classifier (DC), ridge classifier (RC), and bagging classifier (BC), with the hyOPTGB model. The system's efficacy is assessed using various metrics, including accuracy, recall, precision, and F1-score. The hyOPTGB model outperformed the other machine learning models, achieving a 95.3% accuracy rate. The authors also compared the hyOPTGB model against other models proposed by authors who used the same dataset. [ABSTRACT FROM AUTHOR]

Computer vision solutions have become widely used in various industries and as part of daily solutions. One task of computer vision is object detection. With the development of object detection algorithms and the growing number of various kinds of image data, different problems arise in relation to the building of models suitable for various solutions. This paper investigates the influence of parameters used in the training process involved in detecting similar kinds of objects, i.e., the hyperparameters of the algorithm and the training parameters. This experimental investigation focuses on the widely used YOLOv5 algorithm and analyses the performance of different models of YOLOv5 (n, s, m, l, x). In the research, the newly collected construction details (22 categories) dataset is used. Experiments are performed using pre-trained models of the YOLOv5. A total of 185 YOLOv5 models are trained and evaluated. All models are tested on 3300 images photographed on three different backgrounds: mixed, neutral, and white. Additionally, the best-obtained models are evaluated using 150 new images, each of which has several dozen construction details and is photographed against different backgrounds. The deep analysis of different YOLOv5 models and the hyperparameters shows the influence of various parameters when analysing the object detection of similar objects. The best model was obtained when the YOLOv5l was used and the parameters are as follows: coloured images, image size—320; batch size—32; epoch number—300; layers freeze option—10; data augmentation—on; learning rate—0.001; momentum—0.95; and weight decay—0.0007. These results may be useful for various tasks in which small and similar objects are analysed. [ABSTRACT FROM AUTHOR]

Modern active distribution networks (ADNs) witness increasing complexities that require efforts in control practices, including optimal reactive power dispatch (ORPD). Deep reinforcement learning (DRL) is proposed to manage the network's reactive power by coordinating different resources, including distributed energy resources, to enhance performance. However, there is a lack of studies examining DRL elements' performance sensitivity. To this end, in this paper we examine the impact of various DRL reward representations and hyperparameters on the agent's learning performance when solving the ORPD problem for ADNs. We assess the agent's performance regarding accuracy and training time metrics, as well as critic estimate measures. Furthermore, different environmental changes are examined to study the DRL model's scalability by including other resources. Results show that compared to other representations, the complementary reward function exhibits improved performance in terms of power loss minimization and convergence time by 10–15% and 14–18%, respectively. Also, adequate agent performance is observed to be neighboring the best-suited value of each hyperparameter for the studied problem. In addition, scalability analysis depicts that increasing the number of possible action combinations in the action space by approximately nine times results in 1.7 times increase in the training time. [ABSTRACT FROM AUTHOR]

The external and middle ear conditions are diagnosed using a digital otoscope. The clinical diagnosis of ear conditions is suffered from restricted accuracy due to the increased dependency on otolaryngologist expertise, patient complaint, blurring of the otoscopic images, and complexity of lesions definition. There is a high requirement for improved diagnosis algorithms based on otoscopic image processing. This paper presented an ear diagnosis approach based on a convolutional neural network (CNN) as feature extraction and long short-term memory (LSTM) as a classifier algorithm. However, the suggested LSTM model accuracy may be decreased by the omission of a hyperparameter tuning process. Therefore, Bayesian optimization is used for selecting the hyperparameters to improve the results of the LSTM network to obtain a good classification. This study is based on an ear imagery database that consists of four categories: normal, myringosclerosis, earwax plug, and chronic otitis media (COM). This study used 880 otoscopic images divided into 792 training images and 88 testing images to evaluate the approach performance. In this paper, the evaluation metrics of ear condition classification are based on a percentage of accuracy, sensitivity, specificity, and positive predictive value (PPV). The findings yielded a classification accuracy of 100%, a sensitivity of 100%, a specificity of 100%, and a PPV of 100% for the testing database. Finally, the proposed approach shows how to find the best hyperparameters concerning the Bayesian optimization for reliable diagnosis of ear conditions under the consideration of LSTM architecture. This approach demonstrates that CNN-LSTM has higher performance and lower training time than CNN, which has not been used in previous studies for classifying ear diseases. Consequently, the usefulness and reliability of the proposed approach will create an automatic tool for improving the classification and prediction of various ear pathologies. [ABSTRACT FROM AUTHOR]

This paper investigates the effect of Support Vector Regression hyperparameters optimization on electrical load prediction. Accurate and robust load prediction helps policy makers in the energy sector to make inform decision and reduce losses. To achieve this, Bayesian optimization technique was employed for the hyperparameters optimization which are then used for the load prediction. The hyperparameters are the regularization parameters and the epsilon. In addition, the effects of sliding window during the load prediction were also evaluated. The sliding window values were varied from 1 to 5. The results showed that the sliding window of 1 had the optimized hyperparameters with the best performing evaluation metrics of 0.01912 and 0.09493 for MSE and MAE respectively. [ABSTRACT FROM AUTHOR]

The most frequent and noticeable natural calamity in the Karakoram region is landslides. Extreme landslides have occurred frequently along Karakoram Highway, particularly during monsoons, causing a major loss of life and property. Therefore, it is necessary to look for a solution to increase growth and vigilance in order to lessen losses related to landslides caused by natural disasters. By utilizing contemporary technologies, an early warning system might be developed. Artificial neural networks (ANNs) are widely used nowadays across many industries. This paper's major goal is to provide new integrative models for assessing landslide susceptibility in a prone area in the north of Pakistan. To achieve this, the training of an artificial neural network (ANN) was supervised using metaheuristic and Bayesian techniques: Particle Swarm Optimization (PSO) algorithm, Genetic algorithm (GA), Bayesian Optimization Gaussian Process (BO_GP), and Bayesian Optimization Tree-structured Parzen Estimator (BO_TPE). In total, 304 previous landslides and the eight most prevalent conditioning elements were combined to form a geospatial database. The models were hyperparameter optimized, and the best ones were employed to generate susceptibility maps. The obtained area under the curve (AUC) accuracy index demonstrated that the maps produced by both Bayesian and metaheuristic algorithms are highly accurate. The effectiveness and efficiency of applying ANNs for landslide mapping, susceptibility analysis, and forecasting were studied in this research, and it was observed from experimentation that the performance differences for GA, BO_GP, and PSO compared to BO_TPE were relatively small, ranging from 0.32% to 1.84%. This suggests that these techniques achieved comparable performance to BO_TPE in terms of AUC. However, it is important to note that the significance of these differences can vary depending on the specific context and requirements of the ML task. Additionally, in this study, we explore eight feature selection algorithms to determine the geospatial variable importance for landslide susceptibility mapping along the Karakoram Highway (KKH). The algorithms considered include Information Gain, Variance Inflation Factor, OneR Classifier, Subset Evaluators, principal components, Relief Attribute Evaluator, correlation, and Symmetrical Uncertainty. These algorithms enable us to evaluate the relevance and significance of different geospatial variables in predicting landslide susceptibility. By applying these feature selection algorithms, we aim to identify the most influential geospatial variables that contribute to landslide occurrences along the KKH. The algorithms encompass a diverse range of techniques, such as measuring entropy reduction, accounting for attribute bias, generating single rules, evaluating feature subsets, reducing dimensionality, and assessing correlation and information sharing. The findings of this study will provide valuable insights into the critical geospatial variables associated with landslide susceptibility along the KKH. These insights can aid in the development of effective landslide mitigation strategies, infrastructure planning, and targeted hazard management efforts. Additionally, the study contributes to the field of geospatial analysis by showcasing the applicability and effectiveness of various feature selection algorithms in the context of landslide susceptibility mapping. [ABSTRACT FROM AUTHOR]

Cement-based materials are widely used in transportation, construction, national defense, and other fields, due to their excellent properties. High performance, low energy consumption, and environmental protection are essential directions for the sustainable development of cement-based materials. To alleviate the environmental pressure caused by carbon emissions in cement production, this paper studies cement-based materials containing metakaolin by a comparison of prediction models for the compressive strength. To more accurately evaluate the compressive strength of metakaolin cement-based materials, this paper compares the prediction effects of four models, namely, support vector machine (SVM), decision tree (DT), k-nearest neighbor (KNN), and random forest (RF), with hyperparameters optimized by the Firefly Algorithm (FA) to study the compressive strength of cement-based materials containing metakaolin. The results demonstrated that the RF model showed the optimized prediction effect considering the lowest RSME value and the highest R value among the hybrid models for predicting metakaolin cement-based materials' compressive strength. The importance test showed that the cement grade and the water-to-binder ratio greatly influence the compressive strength of cement-based materials with metakaolin compared to the other design parameters. [ABSTRACT FROM AUTHOR]

Compressive sensing is a sub-Nyquist sampling technique for efficient signal acquisition and reconstruction of sparse or compressible signals. In order to account for the sparsity of the underlying signal of interest, it is common to use sparsifying priors such as Bernoulli–Gaussian-inverse Gamma (BGiG) and Gaussian-inverse Gamma (GiG) priors on the components of the signal. With the introduction of variational Bayesian inference, the sparse Bayesian learning (SBL) methods for solving the inverse problem of compressive sensing have received significant interest as the SBL methods become more efficient in terms of execution time. In this paper, we consider the sparse signal recovery problem using compressive sensing and the variational Bayesian (VB) inference framework. More specifically, we consider two widely used Bayesian models of BGiG and GiG for modeling the underlying sparse signal for this problem. Although these two models have been widely used for sparse recovery problems under various signal structures, the question of which model can outperform the other for sparse signal recovery under no specific structure has yet to be fully addressed under the VB inference setting. Here, we study these two models specifically under VB inference in detail, provide some motivating examples regarding the issues in signal reconstruction that may occur under each model, perform comparisons and provide suggestions on how to improve the performance of each model. [ABSTRACT FROM AUTHOR]

In cyber physical system, the machine learning model is a competitive tool and has been successfully applied in this field. However, applying machine learning models to specific problems requires highly experienced manual tuning, which is a process of constant trial and error. Therefore, it often takes huge resources and time when faced with complex problems. In this paper, we propose an efficient method EAT-ML for automatically tuning machine learning models in cyber physical system. This method can automatically tune the hyperparameters of the machine learning model by using a controller with little human intervention. Specifically, the controller sequentially selects the hyperparameters of the machine learning model, and then uses the accuracy obtained on the verification set as the reward value signal. Finally, the PPO algorithm is used to calculate the loss function to update the internal parameters of the controller. In order to further improve the efficiency of the tuning method, we use the previous optimization experience by reconstructing the advantage function. In the experiment, our proposed method can show the best performance on most tasks by comparing other automatic tuning methods. In addition, the effectiveness and feasibility of the various components of the proposed method are also verified through ablation experiments. [ABSTRACT FROM AUTHOR]

Copyright of Science: Future of Lithuania / Mokslas: Lietuvos Ateitis is the property of Vilnius Gediminas Technical University and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

In a deep learning model, the effect of the model may vary depending on the setting of the hyperparameters. Despite the importance of such hyperparameter determination, most previous studies related to burst detection models of the water supply pipe network used hyperparameters applied in other fields as-is or made a trial-and-error setting based on experience, which is a limitation. In this paper, a study was conducted on the deep learning hyperparameter determination of a deep neural network (DNN)-based real-time detection model of pipe burst accidents. The pipe burst model predicted water pressure by using operation data in units of 1 min, and the data period applied for the model training was less than 1 month (1, 2, and 3 weeks) in order to consider frequent changes in the system. A sensitivity analysis was first performed on the type of activation function and the period of the learning data, which may have different effects depending on the characteristics of the target problem. The number of hidden layers related to the network structure and the number of neurons in each hidden layer were set as hyperparameters for additional sensitivity analysis. The sensitivity analysis results were derived and compared using four quantified prediction error indicators. In addition, the model running time was analyzed to evaluate the practical applicability of the development model. From the results, it was confirmed that excellent effects could be expected if using a rectifier function as the activation function, 144 nodes in the hidden layer, which is eight times the number of nodes in the input layer, and four hidden layers. Additionally, by analyzing the appropriate period of training data required for model pressure prediction through prediction error and driving time, it was confirmed that it was most appropriate to use the data of two weeks. By applying the hyperparameter values determined through detailed sensitivity analysis and by applying the data of one week including actual burst accidents to the built-up pressure prediction model, the accident detection and predictive performance of the model were verified. The rational determination of the period of input factors for the optimal hyperparameter setting and model building, as in this study, is very necessary and very important as it can serve to ensure the continuity of the operation effects of the deep learning model. [ABSTRACT FROM AUTHOR]

In this paper, the convolutional neural network is applied to the processing of stylistic features in the images of landscape paintings of Wei Jin and North and South Dynasties. After convolution and pooling activation, rich and less parametric feature maps are generated to lay a solid foundation for generative adversarial networks. Meanwhile, in order to ensure the rigor of the image dataset of landscape painting images under scene transformation, an improvement is proposed on the basis of traditional GAN technology, and a function represents the improved CycleGAN model. The experimental dataset has been constructed, and the loss function, network training, and hyperparameters have been designed for the model. The visual and conversion outcomes of the model-generated images are evaluated through simulation experiments. The migration of all landscape painting style textures is within the range of (0.5,2.4) at a gray level of 8. The migration styles increase as the gray level increases, and the data between groups does not overlap. In the subjective evaluation of the landscape painting style transformation generated by the model used in this paper, the score of the generated image is 3.8135 points, while the number of high scores is more than 65% of the votes. The model constructed in this paper has a better effect on the transformation of the image style of the Wei, Jin, and North and South Dynasties, and it can satisfy people’s visual needs.

The runtime behavior of Simulated Annealing (SA), similar to other metaheuristics, is controlled by hyperparameters. For SA, hyperparameters affect how "temperature" varies over time, and "temperature" in turn affects SA's decisions on whether or not to transition to neighboring states. It is typically necessary to tune the hyperparameters ahead of time. However, there are adaptive annealing schedules that use search feedback to evolve the "temperature" during the search. A classic and generally effective adaptive annealing schedule is the Modified Lam. Although effective, the Modified Lam can be sensitive to the scale of the cost function, and is sometimes slow to converge to its target behavior. In this paper, we present a novel variation of the Modified Lam that we call Self-Tuning Lam, which uses early search feedback to auto-adjust its self-adaptive behavior. Using a variety of discrete and continuous optimization problems, we demonstrate the ability of the Self-Tuning Lam to nearly instantaneously converge to its target behavior independent of the scale of the cost function, as well as its run length. Our implementation is integrated into Chips-n-Salsa, an open-source Java library for parallel and self-adaptive local search. [ABSTRACT FROM AUTHOR]

Considering the dynamics of the economic environment and the amount of data generated every second, the decision-making process is changing and becomes data driven, highly influencing the business strategies setup in order to keep the competitive advantage. However, without technology, data analysis would not be feasible, reason why machine learning is seen as a disruptive innovation for businesses, especially due to its capacity to convert data into actionable outcomes. Though, for a high-quality machine learning model result, algorithm selection and hyperparameters optimization play vital roles, hence became high-interest topics in the field. To achieve this, various automatic selection methods have been proposed and the aim of this paper is to compare two of them - GridSearch and RandomizedSearch - and assess their impact on the model accuracy by comparing with the results obtained when default hyperparameters were applied. [ABSTRACT FROM AUTHOR]

Neural networks have several useful applications in machine learning. However, benefiting from the neural-network architecture can be tricky in some instances due to the large number of parameters that can influence performance. In general, given a particular dataset, a data scientist cannot do much to improve the efficiency of the model. However, by tuning certain hyperparameters, the model's accuracy and time of execution can be improved. Hence, it is of utmost importance to select the optimal values of hyperparameters. Choosing the optimal values of hyperparameters requires experience and mastery of the machine learning paradigm. In this paper, neural network-based architectures are tested based on altering the values of hyperparameters for handwritten-based digit recognition. Various neural network-based models are used to analyze different aspects of the same, primarily accuracy based on hyperparameter values. The extensive experimentation setup in this article should, therefore, provide the most accurate and time-efficient solution models. Such an evaluation will help in selecting the optimized values of hyperparameters for similar tasks. [ABSTRACT FROM AUTHOR]

The paper focuses on the hepatitis C virus (HCV) infection in Egypt, which has one of the highest rates of HCV in the world. The high prevalence is linked to several factors, including the use of injection drugs, poor sterilization practices in medical facilities, and low public awareness. This paper introduces a hyOPTGB model, which employs an optimized gradient boosting (GB) classifier to predict HCV disease in Egypt. The model’s accuracy is enhanced by optimizing hyperparameters with the OPTUNA framework. Min-Max normalization is used as a preprocessing step for scaling the dataset values and using the forward selection (FS) wrapped method to identify essential features. The dataset used in the study contains 1385 instances and 29 features and is available at the UCI machine learning repository. The authors compare the performance of five machine learning models, including decision tree (DT), support vector machine (SVM), dummy classifier (DC), ridge classifier (RC), and bagging classifier (BC), with the hyOPTGB model. The system’s efficacy is assessed using various metrics, including accuracy, recall, precision, and F1-score. The hyOPTGB model outperformed the other machine learning models, achieving a 95.3% accuracy rate. The authors also compared the hyOPTGB model against other models proposed by authors who used the same dataset.

Endoscopy is an important tool for gastric cancer screening. Due to the lack of effective decision support system for endoscopy, the detection of gastric cancer in the clinic is usually with low sensitivity. In this paper, we propose a Genetic Algorithm optimized Neural Network (GAoNN) approach for gastric cancer detection based on endoscopy reports mining. Considering the fact that gastric cancer sensitivity can significantly improve the 5‐year survival rate of patients, both the prediction accuracy and the sensitivity are employed to construct a multiobjective optimization model for enhancing the classification performance of GAoNN. In particular, we extended an effective genetic algorithm Nondominated Sorting Genetic Algorithm II (NSGA‐II) to train a neural network and reduced the complexity in training hyperparameters and improved the efficiency by substituting the computationally intensive stochastic gradient descent (SGD) algorithm in a neural network. Specifically, we designed the novel crossover and mutation operators and modified the nondominated ranking and crowding distance sorting procedures in NSGA‐II for GAoNN. Through testing on 8,546 real‐world endoscopy reports, we show that GAoNN achieves a prediction accuracy up to 83.74%, which is better than several competitors by significantly increasing sensitivity to 83.14%. GAoNN also reduces the training time by 30.94% when compared with conventional SGD‐based training, which indicates the feasibility of GAoNN in clinical practice. [ABSTRACT FROM AUTHOR]

Human intelligence is constituted by a multitude of cognitive functions activated either directly or indirectly by external stimuli of various kinds. Computational approaches to the cognitive sciences and to neuroscience are partly premised on the idea that computational simulations of such cognitive functions and brain operations suspected to correspond to them can help to further uncover knowledge about those functions and operations, specifically, how they might work together. These approaches are also partly premised on the idea that empirical neuroscience research, whether following on from such a simulation (as indeed simulation and empirical research are complementary) or otherwise, could help us build better artificially intelligent systems. This is based on the assumption that principles by which the brain seemingly operate, to the extent that it can be understood as computational, should at least be tested as principles for the operation of artificial systems. This paper explores some of the principles of the brain that seem to be responsible for its autonomous, problem-adaptive nature. The brain operating system (BrainOS) explicated here is an introduction to ongoing work aiming to create a robust, integrated model, combining the connectionist paradigm underlying neural networks and the symbolic paradigm underlying much else of AI. BrainOS is an automatic approach that selects the most appropriate model based on the (a) input at hand, (b) prior experience (a history of results of prior problem solving attempts), and (c) world knowledge (represented in the symbolic way and used as a means to explain its approach). It is able to accept diverse and mixed input data types, process histories and objectives, extract knowledge and infer a situational context. BrainOS is designed to be efficient through its ability to not only choose the most suitable learning model but to effectively calibrate it based on the task at hand. [ABSTRACT FROM AUTHOR]

A detailed parametric analysis is presented, where the recent method based on the reservoir computing paradigm, including its statistical robustness, is studied. It is observed that the prediction capabilities of the reservoir computing approach strongly depend on the random initialization of both the input and the reservoir layers. Special emphasis is put on finding the region in the hyperparameter space where the ensemble-averaged training and generalization errors together with their variance are minimized. The statistical analysis presented here is based on the projection on proper elements method. [ABSTRACT FROM AUTHOR]

The diagnosis of different pathologies and stages of cancer using whole histopathology slide images (WSI) is the gold standard for determining the degree of tissue metastasis. The use of deep learning systems in the field of medical images, especially histopathology images, is becoming increasingly important. The training and optimization of deep neural network models involve fine-tuning parameters and hyperparameters such as learning rate, batch size (BS), and boost to improve the performance of the model in task-specific applications. Tuning hyperparameters is a major challenge in designing deep neural network models, having a large impact on the performance. This paper analyzes how the parameters and hyperparameters of a deep learning architecture affect the classification of colorectal cancer (CRC) histopathology images using the well-known VGG19 model. This paper also discusses the pre-processing of these images, such as the use of color normalization and stretching transformations on the data set. Among these hyperparameters, the most important neural network hyperparameter is the learning rate (LR). In this paper, different strategies for the optimization of LR are analyzed (both static and dynamic) and a new experiment based on the variation of LR is proposed (the relevance of dynamic strategies over fixed LR is highlighted), after each layer of the neural network together with decreasing variations according to the epochs. The results obtained are very remarkable, obtaining in the simulation an accurate system that achieves 96.4% accuracy on test images (for nine different tissue classes) using the triangular-cyclic learning rate.

This paper demonstrates the differences between popular transformation-based input representations for vibration-based machine fault diagnosis. This paper highlights the dependency of different input representations on hyperparameter selection with the results of training different configurations of classical convolutional neural networks (CNNs) with three common benchmarking datasets. Raw temporal measurement, Fourier spectrum, envelope spectrum, and spectrogram input types are individually used to train CNNs. Many configurations of CNNs are trained, with variable input sizes, convolutional kernel sizes and stride. The results show that each input type favors different combinations of hyperparameters, and that each of the datasets studied yield different performance characteristics. The input sizes are found to be the most significant determiner of whether overfitting will occur. It is demonstrated that CNNs trained with spectrograms are less dependent on hyperparameter optimization over all three datasets. This paper demonstrates the wide range of performance achieved by CNNs when preprocessing method and hyperparameters are varied as well as their complex interaction, providing researchers with useful background information and a starting place for further optimization.

One of the biggest challenge in the field of deep learning is the parameter selection and optimization process. In recent years different algorithms have been proposed including bio-inspired solutions to solve this problem, however, there are many challenges including local minima, saddle points, and vanishing gradients. In this paper, we introduce the Whale Optimisation Algorithm (WOA) based on the swarm foraging behavior of humpback whales to optimise neural network hyperparameters. We wish to stress that to the best of our knowledge this is the first attempt that uses Whale Optimisation Algorithm for the optimisation task of hyperparameters. After a detailed description of the WOA algorithm we formulate and explain the application in deep learning, present the implementation, and compare the proposed algorithm with other well-known algorithms including widely used Grid and Random Search methods. Additionally, we have implemented a third dimension feature analysis to the original WOA algorithm to utilize 3D search space (3D-WOA). Simulations show that the proposed algorithm can be successfully used for hyperparameters optimization, achieving accuracy of 89.85% and 80.60% for Fashion MNIST and Reuters datasets, respectively. [ABSTRACT FROM AUTHOR]

The present paper considers an open problem of setting hyperparameters for convolutional neural networks aimed at image classification. Since selecting filter spatial extents for convolutional layers is a topical problem, it is approximately solved by accumulating statistics of the neural network performance. The network architecture is taken on the basis of the MNIST database experience. The eight-layered architecture having four convolutional layers is nearly best suitable for classifying small and medium size images. Image databases are formed of grayscale images whose size range is 28 × 28 to 64 × 64 by step 2. Except for the filter spatial extents, the rest of those eight layer hyperparameters are unalterable, and they are chosen scrupulously based on rules of thumb. A sequence of possible filter spatial extents is generated for each size. Then sets of four filter spatial extents producing the best performance are extracted. The rule of this extraction that allows selecting the best filter spatial extents is formalized with two conditions. Mainly, difference between maximal and minimal extents must be as minimal as possible. No unit filter spatial extent is recommended. The secondary condition is that the filter spatial extents should constitute a non-increasing set. Validation on MNIST and CIFAR- 10 databases justifies such a solution, which can be extended for building convolutional neural network classifiers of colour and larger images. [ABSTRACT FROM AUTHOR]

In India, approximately one-third of agricultural produce is wasted every year due to the different issues present in the post-harvest supply chain stages. The supply chain management of perishable products becomes complex and challenging due to the inclusion of its perishability dynamics. Quality is the main factor that governs the buying or discarding of perishable products by consumers. Therefore, the main aim of this research work is to develop an accurate and efficient image processing model for the classification of the product (Tomato) based on its quality for managing the supply chain. There are three novelties in this research work. A four-stage supply chain architecture integrated with the image processing system at mandi, and warehouses is proposed (First). This image processing system is developed in two stages. In stage I, the acquired images of tomato during its life cycle are labelled with the help of machine learning algorithms (Second). This labelled data is used in stage II for the development of a classification model to segregate the product into various grades. For this, an optimized architecture of seven-layer Convolutional Neural Network (CNN) model is developed followed by optimization of its hyperparameters simultaneously using Design of Experiments (DOE) technique (Third). The optimized CNN model achieved maximum accuracy of 88.40% and reported an execution time of 7 min. Further, the results of standard hyperparameter optimization techniques like Grid search, Random search, Bayesian, and Hyperband are compared with the proposed DOE technique on the optimized CNN architecture. The work done in this paper enables the supply chain managers to take accurate and rapid decisions for pricing, procurement, storage, and transportation at various stages of the supply chain leading to Industry 4.0. This will result in reduced post-harvest losses and simultaneously achieve the benefits across social, economic, and environmental dimensions of sustainability leading to better supply chain management. • A supply chain architecture integrated with image processing system is proposed. • Decisions related to pricing, procurement, storage and transportation can be taken. • The quality of product is modelled and labelled using Machine Learning algorithms. • A CNN model is developed for quality-based classification and optimized using DoE. • A comparison of DoE with Grid search, Bayesian and Hyperband is presented. [ABSTRACT FROM AUTHOR]

Featured Application: Python application that uses data science and machine learning to estimate the main properties of acid tars. Its main advantage is that determinations for acid tar properties are no longer necessary, thus saving time and money. However, good machine learning estimations are highly dependent on the number and quality of the training data, meaning that the larger and more consistent the training database, the better the estimations. Hazardous petroleum wastes are an inevitable source of environmental pollution. Leachates from these wastes could contaminate soil and potable water sources and affect human health. The management of acid tars, as a byproduct of refining and petrochemical processes, represented one of the major hazardous waste problems in Romania. Acid tars are hazardous and toxic waste and have the potential to cause pollution and environmental damage. The need for the identification, study, characterization, and subsequently either the treatment, valorization, or elimination of acid tars is determined by the fact that they also have high concentrations of hydrocarbons and heavy metals, toxic for the storage site and its neighboring residential area. When soil contamination with acid tars occurs, sustainable remediation techniques are needed to restore soil quality to a healthy production state. Therefore, it is necessary to ensure a rapid but robust characterization of the degree of contamination with hydrocarbons and heavy metals in acid tars so that appropriate techniques can then be used for treatment/remediation. The first stage in treating these acid tars is to determine its properties. This article presents a software program that uses machine learning to estimate selected properties of acid tars (pH, Total Petroleum Hydrocarbons—TPH, and heavy metals). The program uses the Automatic Machine Learning technique to determine the Machine Learning algorithm that has the lowest estimation error for the given dataset, with respect to the Mean Average Error and Root Mean Squared Error. The chosen algorithm is used further for properties estimation, using the R2 correlation coefficient as a performance criterion. The dataset used for training has 82 experimental points with continuous, unique values containing the coordinates and depth of acid tar samples and their properties. Based on an exhaustive search performed by the authors, a similar study that considers machine learning applications was not found in the literature. Further research is required because the method presented therein can be improved because it is dataset dependent, as is the case with every ML problem. [ABSTRACT FROM AUTHOR]

In road traffic management, high-speed vehicle detection is often affected by factors such as vehicle speed, weather, camera angle, and image resolution, making vehicle detection on express roads very challenging, Therefore, we propose a Taguchi-based You Only Look Once (YOLOv7) model, called T-YOLOv7, for vehicle detection on high-speed roads. The Taguchi method is used to optimize the combination of hyperparameters of YOLOv7. Experimental results show the precision rate, recall rate, and F1-score of the proposed T-YOLOv7 to be 82.2, 86.3, and 84.2%, respectively. Compared with the original YOLOv7, T-YOLOv7 has improved precision, recall, and F1-score by 12.1, 17.9, and 15.0 percentage points, respectively. Compared with YOLOv4, the improvements in precision, recall, and F1-score using T-YOLOv7 are 4.0, 2.4, and 3.3 percentage points, respectively. The experimental results also show that the proposed T-YOLOv7 is effective in adjusting hyperparameters through the Taguchi method and can be applied to realtime vehicle detection in real environments. [ABSTRACT FROM AUTHOR]

In recent years, different variants of the botnet are targeting government, private organizations and there is a crucial need to develop a robust framework for securing the IoT (Internet of Things) network. In this paper, a Hadoop based framework is proposed to identify the malicious IoT traffic using a modified Tomek-link under-sampling integrated with automated Hyper-parameter tuning of machine learning classifiers. The novelty of this paper is to utilize a big data platform for benchmark IoT datasets to minimize computational time. The IoT benchmark datasets are loaded in the Hadoop Distributed File System (HDFS) environment. Three machine learning approaches namely naive Bayes (NB), K-nearest neighbor (KNN), and support vector machine (SVM) are used for categorizing IoT traffic. Artificial immune network optimization is deployed during cross-validation to obtain the best classifier parameters. Experimental analysis is performed on the Hadoop platform. The average accuracy of 99% and 90% is obtained for BoT_IoT and ToN_IoT datasets. The accuracy difference in ToN-IoT dataset is due to the huge number of data samples captured at the edge layer and fog layer. However, in BoT-IoT dataset only 5% of the training and test samples from the complete dataset are considered for experimental analysis as released by the dataset developers. The overall accuracy is improved by 19% in comparison with state-of-the-art techniques. The computational times for the huge datasets are reduced by 3–4 hours through Map Reduce in HDFS. [ABSTRACT FROM AUTHOR]

Bug tracking systems use repositories to keep track of the bugs to improve software quality. A manual analysis of each bug and classifying it according to its severity is an unmanageable job. Therefore, it is imperative to correctly classify the severity of a bug, which otherwise might be misclassified by a laymen user. Text mining techniques have the potential to analyze such massive databases of the textual description of bug reports to classify bug severity levels adequately. Word embedding is a state-of-art text mining technique that captures the semantics of the text and group the words according to their relevance in a document. Words are embedded into real-valued vectors through word-embedding models. Word2vec is a word embedding model, proven to be effective in representing word meanings. However, the configuration of hyperparameters and feature selection affects the performance of word2vec. Tuning hyperparameter is a time-consuming process, and it is crucial to identify the correct set of parameters. The paper outlines the effectiveness of word2vec technique on the efficiency of classifiers to predict a bug severity from a bug report and examines the impact of different averaging methods, including the configuration of word embedding parameters on the classifiers' efficiency. Results show that the bigger window size enhances the performance of classifiers; however, the influence of the minimum word count parameter was found to be mixed and depends on the selected data sets. Further, out of the classifiers used, Random Forest and Xgboost could classify the severity level for classes with few records or a rare occurrence of words specific to each class. Otherwise, Support Vector Machine and Naive Bayes classifiers performed better and worst, respectively. [ABSTRACT FROM AUTHOR]

Cement-based materials are widely used in transportation, construction, national defense, and other fields, due to their excellent properties. High performance, low energy consumption, and environmental protection are essential directions for the sustainable development of cement-based materials. To alleviate the environmental pressure caused by carbon emissions in cement production, this paper studies cement-based materials containing metakaolin by a comparison of prediction models for the compressive strength. To more accurately evaluate the compressive strength of metakaolin cement-based materials, this paper compares the prediction effects of four models, namely, support vector machine (SVM), decision tree (DT), k-nearest neighbor (KNN), and random forest (RF), with hyperparameters optimized by the Firefly Algorithm (FA) to study the compressive strength of cement-based materials containing metakaolin. The results demonstrated that the RF model showed the optimized prediction effect considering the lowest RSME value and the highest R value among the hybrid models for predicting metakaolin cement-based materials’ compressive strength. The importance test showed that the cement grade and the water-to-binder ratio greatly influence the compressive strength of cement-based materials with metakaolin compared to the other design parameters.

The current approach to diagnosing and classifying brain tumors relies on the histological evaluation of biopsy samples, which is invasive, time-consuming, and susceptible to manual errors. These limitations underscore the pressing need for a fully automated, deep-learning-based multi-classification system for brain malignancies. This article aims to leverage a deep convolutional neural network (CNN) to enhance early detection and presents three distinct CNN models designed for different types of classification tasks. The first CNN model achieves an impressive detection accuracy of 99.53% for brain tumors. The second CNN model, with an accuracy of 93.81%, proficiently categorizes brain tumors into five distinct types: normal, glioma, meningioma, pituitary, and metastatic. Furthermore, the third CNN model demonstrates an accuracy of 98.56% in accurately classifying brain tumors into their different grades. To ensure optimal performance, a grid search optimization approach is employed to automatically fine-tune all the relevant hyperparameters of the CNN models. The utilization of large, publicly accessible clinical datasets results in robust and reliable classification outcomes. This article conducts a comprehensive comparison of the proposed models against classical models, such as AlexNet, DenseNet121, ResNet-101, VGG-19, and GoogleNet, reaffirming the superiority of the deep CNN-based approach in advancing the field of brain tumor classification and early detection. [ABSTRACT FROM AUTHOR]

Traditionally Genetic algorithms are thought of as brute force approaches, aimed to arrive at solutions to problems which do not have a specific answer. In problems where the data is not structured for the general implementation of a specific idea, genetic algorithms are most useful. This paper proposes to mitigate the above problem of brute force approaches through elucidation of procedures ranging from exploratory analysis, followed by pattern analysis and classification. This novel conceptualization of the scheme and design will help in arriving at solutions through reduced iterations. Research conducted involves dropping of poorly performing hypotheses, controlled mutation, thereby adding a dimension of intelligence to evolutionary algorithms. The following paper describes the methodology used to solve the problem of addition of numbers using evolutionary algorithms of Neural Networks, whilst building intelligence into the system. The specific problem of addition has been dealt with in the following paper, however the same design philosophy can be extended for a paraphernalia of problems. The end goal is to obtain a generation of adroit and capable hypotheses to solve the problem in reduced number of iterations. The solution provided is generic and can be reused, it has been applied to a specific problem in the following paper.

Background: Heart disease is one of the most recurrent and worrying health problems today, due to its multiple complications, including: stroke, cardiac arrest, retinopathy, etc. Objective: Propose a method and 4 Stacking models based on hyperparameters to diagnose heart disease. In addition, a web interface was developed with the best model proposed in this study. Methods: First, the dataset used was from the Heart Disease Cleveland ICU, which was 918 patient records and 12 attributes. Therefore, the paper was composed of the following stages: Cleaning and Pre-processing; Describe data; Training and testing data; Cross validation; Calibration of models; and modelling and evaluation, also compare the different techniques proposed to predict heart disease using Stacking ensemble based on hyperparameters taking into account the performance evaluation parameters. Results: Stacking 1 (Logistic regression) with oversampling and AdaBoost-SVM with hyperparameter in the test obtained higher Accuracy (88.24%), and ROC Curve (92.00%), while too Stacking 1 (Logistic regression) with oversampling reached a better Precision (88.54%), but the AdaBoost-SVM algorithm using hyperparameter achieved a high value of Sensitivity (88.14%) and F1-Score (88.19%). Conclusions: Implementing 4 stacking models based on hyperparameters, it helps to make an early diagnosis of heart disease and greater precision, and decrease the quantity of deceases caused by it. Therefore, by using the combined method, an improvement in heart disease prediction was observed, surpassing the performance of the independent algorithms used.

Broadband power line communication based on an orthogonal frequency division multiplexing scheme is considered in this paper. In addition to multipath fading and frequency selectivity, the power line communication channel is influenced by impulsive noise. More bit errors in power line communications are caused by these effects. In this paper, a relevance vector machine (RVM) is applied to the received complex data in order to estimate the power line communication channel impulse response in a baseband model. The bit error rate and mean squared error (MSE) performances under the impulsive noise and multipath effects are analyzed. It is shown that introducing a new kernel can increase the robustness of the proposed method against the critical effects of impulsive noise and multipath with low complexity. It is also shown that applying proper values to hyperparameters based on the minimum mean square error criterion can increase convergence speed and decrease MSE in our proposed method. Experimental results show the better performance of proposed algorithm compared to recently reported methods such as the Huang method and conventional RVM with Gaussian kernel function in complex mode, namely complex RVM. [ABSTRACT FROM AUTHOR]

Over the past decades, drones have become more attainable by the public due to their widespread availability at affordable prices. Nevertheless, this situation sparks serious concerns in both the cyber and physical security domains, as drones can be employed for malicious activities with public safety threats. However, detecting drones instantly and efficiently is a very difficult task due to their tiny size and swift flights. This paper presents a novel drone detection method using deep convolutional learning and deep transfer learning. The proposed algorithm employs a new feature extraction network, which is added to the modified YOU ONLY LOOK ONCE version2 (YOLOv2) network. The feature extraction model uses bypass connections to learn features from the training sets and solves the “vanishing gradient” problem caused by the increasing depth of the network. The structure of YOLOv2 is modified by replacing the rectified linear unit (relu) with a leaky-relu activation function and adding an extra convolutional layer with a stride of 2 to improve the small object detection accuracy. Using leaky-relu solves the “dying relu” problem. The additional convolution layer with a stride of 2 reduces the spatial dimensions of the feature maps and helps the network to focus on larger contextual information while still preserving the ability to detect small objects. The model is trained with a custom dataset that contains various types of drones, airplanes, birds, and helicopters under various weather conditions. The proposed model demonstrates a notable performance, achieving an accuracy of 77% on the test images with only 5 million learnable parameters in contrast to the Darknet53 + YOLOv3 model, which exhibits a 54% accuracy on the same test set despite employing 62 million learnable parameters.

The large number of images in the different areas and the possibilities of technologies lead to various solutions in automatization using image data. In this paper, tattoo detection and identification were analyzed. The combination of YOLOv5 object detection methods and similarity measures was investigated. During the experimental research, various parameters have been investigated to determine the best combination of parameters for tattoo detection. In this case, the influence of data augmentation parameters, the size of the YOLOv5 models (n, s, m, l, x), and the three main hyperparameters of YOLOv5 were analyzed. Also, the efficiency of the most popular similarity distances cosine and Euclidean was analyzed in the tattoo identification process with the purpose of matching the detected tattoo with the person’s tattoo in the database. Experiments have been performed using the deMSI dataset, where images were manually labeled to be suitable for use by the YOLOv5 algorithm. To validate the results obtained, the newly collected tattoo dataset was used. The results have shown that the highest average accuracy of all tattoo detection experiments has been obtained using the YOLOv5l model, where mAP@0.5:0.95 is equal to 0.60, and mAP@0.5 is equal to 0.79. The accuracy for tattoo identification reaches 0.98, and the F-score is up to 0.52 when the highest cosine similarity tattoo is associated. Meanwhile, to ensure that no suspects will be missed, the cosine similarity threshold value of 0.15 should be applied. Then, photos with higher similarity scores should be analyzed only. This would lead to a 1.0 recall and would reduce the manual tattoo comparison by 20%.

Long Short Term Memory (LSTM) is among the most popular deep learning models used today. It is also being applied to time series prediction which is a particularly hard problem to solve due to the presence of long term trend, seasonal and cyclical fluctuations and random noise. The performance of LSTM is highly dependent on choice of several hyper-parameters which need to be chosen very carefully, in order to get good results. Being a relatively new model, there are no established guidelines for configuring LSTM. In this paper this research gap was addressed. A dataset was created from the Indian stock market and an LSTM model was developed for it. It was then optimized by comparing stateless and stateful models and by tuning for the number of hidden layers. [ABSTRACT FROM AUTHOR]

Breast cancer is the second leading cause of mortality among women. Early and accurate detection plays a crucial role in lowering its mortality rate. Timely detection and classification of breast cancer enable the most effective treatment. Convolutional neural networks (CNNs) have significantly improved the accuracy of tumor detection and classification in medical imaging compared to traditional methods. This study proposes a comprehensive classification technique for identifying breast cancer, utilizing a synthesized CNN, an enhanced optimization algorithm, and transfer learning. The primary goal is to assist radiologists in rapidly identifying anomalies. To overcome inherent limitations, we modified the Ant Colony Optimization (ACO) technique with opposition-based learning (OBL). The Enhanced Ant Colony Optimization (EACO) methodology was then employed to determine the optimal hyperparameter values for the CNN architecture. Our proposed framework combines the Residual Network-101 (ResNet101) CNN architecture with the EACO algorithm, resulting in a new model dubbed EACO–ResNet101. Experimental analysis was conducted on the MIAS and DDSM (CBIS-DDSM) mammographic datasets. Compared to conventional methods, our proposed model achieved an impressive accuracy of 98.63%, sensitivity of 98.76%, and specificity of 98.89% on the CBIS-DDSM dataset. On the MIAS dataset, the proposed model achieved a classification accuracy of 99.15%, a sensitivity of 97.86%, and a specificity of 98.88%. These results demonstrate the superiority of the proposed EACO–ResNet101 over current methodologies. [ABSTRACT FROM AUTHOR]

Deep neural networks are widely used in the field of image processing for micromachines, such as in 3D shape detection in microelectronic high-speed dispensing and object detection in microrobots. It is already known that hyperparameters and their interactions impact neural network model performance. Taking advantage of the mathematical correlations between hyperparameters and the corresponding deep learning model to adjust hyperparameters intelligently is the key to obtaining an optimal solution from a deep neural network model. Leveraging these correlations is also significant for unlocking the "black box" of deep learning by revealing the mechanism of its mathematical principle. However, there is no complete system for studying the combination of mathematical derivation and experimental verification methods to quantify the impacts of hyperparameters on the performances of deep learning models. Therefore, in this paper, the authors analyzed the mathematical relationships among four hyperparameters: the learning rate, batch size, dropout rate, and convolution kernel size. A generalized multiparameter mathematical correlation model was also established, which showed that the interaction between these hyperparameters played an important role in the neural network's performance. Different experiments were verified by running convolutional neural network algorithms to validate the proposal on the MNIST dataset. Notably, this research can help establish a universal multiparameter mathematical correlation model to guide the deep learning parameter adjustment process. [ABSTRACT FROM AUTHOR]

Computer vision solutions have become widely used in various industries and as part of daily solutions. One task of computer vision is object detection. With the development of object detection algorithms and the growing number of various kinds of image data, different problems arise in relation to the building of models suitable for various solutions. This paper investigates the influence of parameters used in the training process involved in detecting similar kinds of objects, i.e., the hyperparameters of the algorithm and the training parameters. This experimental investigation focuses on the widely used YOLOv5 algorithm and analyses the performance of different models of YOLOv5 (n, s, m, l, x). In the research, the newly collected construction details (22 categories) dataset is used. Experiments are performed using pre-trained models of the YOLOv5. A total of 185 YOLOv5 models are trained and evaluated. All models are tested on 3300 images photographed on three different backgrounds: mixed, neutral, and white. Additionally, the best-obtained models are evaluated using 150 new images, each of which has several dozen construction details and is photographed against different backgrounds. The deep analysis of different YOLOv5 models and the hyperparameters shows the influence of various parameters when analysing the object detection of similar objects. The best model was obtained when the YOLOv5l was used and the parameters are as follows: coloured images, image size—320; batch size—32; epoch number—300; layers freeze option—10; data augmentation—on; learning rate—0.001; momentum—0.95; and weight decay—0.0007. These results may be useful for various tasks in which small and similar objects are analysed.

Modern supervised machine learning algorithms involve hyperparameters that have to be set before running them. Options for setting hyperparameters are default values from the software package, manual configuration by the user or configuring them for optimal predictive performance by a tuning procedure. The goal of this paper is two-fold. Firstly, we formalize the problem of tuning from a statistical point of view, define data-based defaults and suggest general measures quantifying the tunability of hyperparameters of algorithms. Secondly, we conduct a large-scale benchmarking study based on 38 datasets from the OpenML platform and six common machine learning algorithms. We apply our measures to assess the tunability of their parameters. Our results yield default values for hyperparameters and enable users to decide whether it is worth conducting a possibly time consuming tuning strategy, to focus on the most important hyperparameters and to choose adequate hyperparameter spaces for tuning. [ABSTRACT FROM AUTHOR]

Most machine learning algorithms possess hyperparameters. For example, an artificial neural network requires the determination of the number of hidden layers, nodes, and many other parameters related to the model fitting process. Despite this, there is still no clear consensus on how to tune them. The most popular methodology is an exhaustive grid search, which can be highly inefficient and sometimes infeasible. Another common solution is to change one hyperparameter at a time and measure its effect on the model’s performance. However, this can also be inefficient and does not guarantee optimal results since it ignores interactions between the hyperparameters. In this paper, we propose to use the Design of Experiments (DOE) methodology (factorial designs) for screening and Response Surface Methodology (RSM) to tune a machine learning algorithm’s hyperparameters. An application of our methodology is presented with a detailed discussion of the results of a random forest case-study using a publicly available dataset. Benefits include fewer training runs, better parameter selection, and a disciplined approach based on statistical theory. [ABSTRACT FROM AUTHOR]

In recent years, neural network based image priors have been shown to be highly effective for linear inverse problems, often significantly outperforming conventional methods that are based on sparsity and related notions. While pre-trained generative models are perhaps the most common, it has additionally been shown that even untrained neural networks can serve as excellent priors in various imaging applications. In this paper, we seek to broaden the applicability and understanding of untrained neural network priors by investigating the interaction between architecture selection, measurement models (e.g., inpainting vs. denoising vs. compressive sensing), and signal types (e.g., smooth vs. erratic). We motivate the problem via statistical learning theory, and provide two practical algorithms for tuning architectural hyperparameters. Using experimental evaluations, we demonstrate that the optimal hyperparameters may vary significantly between tasks and can exhibit large performance gaps when tuned for the wrong task. In addition, we investigate which hyperparameters tend to be more important, and which are robust to deviations from the optimum. [ABSTRACT FROM AUTHOR]

This work proposes a metaheuristic-based approach to hyperparameter selection in a multilayer perceptron to classify EMG signals. The main goal of the study is to improve the performance of the model by optimizing four important hyperparameters: the number of neurons, the learning rate, the epochs, and the training batches. The approach proposed in this work shows that hyperparameter optimization using particle swarm optimization and the gray wolf optimizer significantly improves the performance of a multilayer perceptron in classifying EMG motion signals. The final model achieves an average classification rate of 93% for the validation phase. The results obtained are promising and suggest that the proposed approach may be helpful for the optimization of deep learning models in other signal processing applications. [ABSTRACT FROM AUTHOR]

Cryptocurrencies are rapidly growing and are increasingly accepted by major commercial vendors. However, along with their rising popularity, they have also become the go-to currency for illicit activities driven by the anonymity they provide. Cryptocurrencies such as the one on the Ethereum blockchain provide a way for entities to hide their real-world identities behind pseudonyms, also known as addresses. Hence, the purpose of this work is to uncover the level of anonymity in Ethereum by investigating multiclass classification models for Externally Owned Accounts (EOAs) of Ethereum. The researchers aim to achieve this by examining patterns of transaction activity associated with these addresses. Using a labelled Ethereum address dataset from Kaggle and the Ethereum crypto dataset by Google BigQuery, an address profiles dataset was compiled based on the transaction history of the addresses. The compiled dataset, consisting of 4371 samples, was used to tune and evaluate the Random Forest, Gradient Boosting and XGBoost classifier for predicting the category of the addresses. The best-performing model found for the problem was the XGBoost classifier, achieving an accuracy of 75.3% with a macro-averaged F1-Score of 0.689. Following closely was the Random Forest classifier, with an accuracy of 73.7% and a macro-averaged F1-Score of 0.641. Gradient Boosting came in last with 73% accuracy and a macro-averaged F1-Score of 0.659. Owing to the data limitations in this study, the overall scores of the best model were weaker in comparison to similar research, with the exception of precision, which scored slightly higher. Nevertheless, the results proved that it is possible to predict the category of an Ethereum wallet address such as Phish/Hack, Scamming, Exchange and ICO wallets based on its transaction behaviour. [ABSTRACT FROM AUTHOR]

Heart disease, globally recognized as a leading cause of death, has seen its impact magnified by the emergence of COVID-19. The heightened demand for early detection and diagnosis of heart disease has forced the development of innovative, intelligent systems. This research offers a novel approach by leveraging extended short-term memory networks (LSTM) and including COVID-19 as a significant parameter in cardiac arrest analysis. A comparative study is conducted between LSTM and other prevalent techniques, such as support vector machines (SVM), linear regression (LR), and artificial neural networks (ANN), focusing on accuracy and other prognostic criteria for heart disease. We aim to develop an intelligent system powered by LSTM to predict heart disease, thereby assisting healthcare professionals in making well-informed decisions about heart disease management, stroke prevention, and patient monitoring. Additionally, hyperparameter tuning has been performed to optimize the LSTM model's performance in cardiac arrest prediction. The results underscore that LSTM, especially when trained with COVID-19 as an input parameter, surpasses other established techniques in prediction accuracy. The proposed model underwent experimental testing, showcasing its proficiency in predicting cardiovascular disease. [ABSTRACT FROM AUTHOR]

There has been a fast deployment of wireless networks in recent years, which has been accompanied by significant impacts on the environment. Among the solutions that have been proven to be effective in reducing the energy consumption of wireless networks is the use of machine learning algorithms in cell traffic management. However, despite promising results, it should be noted that the computations required by machine learning algorithms have increased at an exponential rate. Massive computing has a surprisingly large carbon footprint, which could affect its real-world deployment. Thus, additional attention needs to be paid to the design and parameterization of these algorithms applied in order to reduce the energy consumption of wireless networks. In this article, we analyze the impact of hyperparameters on the energy consumption and performance of machine learning algorithms used for cell traffic prediction. For each hyperparameter (number of layers, number of neurons per layer, optimizer algorithm, batch size, and dropout) we identified a set of feasible values. Then, for each combination of hyperparameters, we trained our model and analyzed energy consumption and the resulting performance. The results from this study reveal a great correlation between hyperparameters and energy consumption, confirming the paramount importance of selecting optimal hyperparameters. A tradeoff between the minimization of energy consumption and the maximization of machine learning performance is suggested. [ABSTRACT FROM AUTHOR]

Network intrusion detection is a relevant cybersecurity research field. The growing number of intrusions requires more sophisticated methods to protect computer networks. Various machine learning algorithms are used to detect network intrusions and anomalies, but their accuracy is limited. In this research, we address the problem of improving network-level intrusion detection by applying hybrid machine-learning algorithms. The paper proposes three new hybrid machine learning methods and investigates their accuracy using two publicly available datasets CSE-CIC-IDS2018 and NSW-NB-15. In order to increase the accuracy of the classification models, hyperparameter optimization was performed. The iteration method and the Chi-square χ2 test were used to identify significant features of the data set. Analyzing the research results, it was found that the highest network anomaly recognition accuracy of 99.34% was achieved by applying a hybrid algorithm consisting of a decision tree, naive Bayesian, and multilayer perceptron algorithms. Achieved result is 3.13% higher than the best accuracy achieved by individual machine learning algorithms. In order to comprehensively evaluate the studied machine learning algorithms and their suitability for detecting intrusions in a computer network, the algorithms were ranked using the SCR, DR, FR ranking methods. Article in Lithuanian. Įsilaužimų aptikimas kompiuterių tinkluose taikant hibridinius mašininio mokymosi metodus Santrauka Viena iš aktualių kibernetinės saugos tyrimų krypčių – tai įsilaužimų arba anomalijų aptikimas kompiuterių tinkle. Įsilaužimų skaičius nuolat didėja, o taikomos įsilaužimo technikos ir metodai sudėtingėja, todėl siekiant apsaugoti kompiuterių tinklą, reikia taikyti vis sudėtingesnius apsaugos metodus. Tinklo įsilaužimams ir anomalijoms nustatyti taikomi įvairūs mašininio mokymosi algoritmai, tačiau jų tikslumas yra ribotas. Siekiama pagerinti tinklo anomalijų aptikimą, taikomi hibridiniai mašininio mokymosi algoritmai. Straipsnyje pasiūlyti trys nauji hibridiniai mašininio mokymosi algoritmai, ištirtas jų tikslumas naudojant du viešai prieinamus duomenų rinkinius, t. y. CSE-CIC-IDS2018 ir NSW-NB-15. Siekiant padidinti klasifikavimo modelių tikslumą, buvo atliktas hiperparametrų optimizavimas. Reikšmingiems duomenų rinkinio požymiams nustatyti taikytas iteracijų metodas ir Chi kvadrato χ2 testas. Analizuojant tyrimo rezultatus, nustatyta, kad aukščiausias tinklo anomalijų atpažinimo tikslumas 99,34 % buvo pasiektas taikant hibridinį algoritmą, sudarytą iš sprendimų medžio, naivaus Bajeso ir daugiasluoksnio perceptrono algoritmų rinkinio. Šis rezultatas yra 3,13 % geresnis, lyginant su geriausiu tikslumu, gautu taikant atskirus mašininio mokymosi algoritmus. Siekiant kompleksiškai įvertinti tirtus mašininio mokymosi algoritmus ir jų tinkamumą įsilaužimams kompiuterių tinkle aptikti, algoritmai buvo sureitinguoti taikant SCR, DR, FR reitingavimo metodus. Reikšminiai žodžiai: tinklo anomalijos, mašininis mokymasis, χ2 Chi kvadratu testas, hiperparametrai, hibridiniai algoritmai.