Unsupervised Clustering and Supervised Machine Learning for Lightning Classification: Application to Identifying EIPs for Ground-based TGF Detection

<p>Terrestrial Gamma-ray Flashes (TGFs) are short emissions of high-energy photons associated with thunderstorms. Since their discovery, it has been clear that they are associated with lightning, and several case studies have shown that the TGFs are produced in the initial phase of the lightning flash. However, it has not been tested whether this is true in general. Here we present such study using the largest TGF sample available to date from the RHESSI, Fermi, AGILE, ASIM catalogs, combined with ground-based radio lightning detection data. Based on stacking analysis of the TGFs and associated lightning activity, together with the high temporal resolution of the optical data from the ASIM photometers, we show that, indeed, TGFs are produced at the beginning of lightning flashes. We also find that the detected sferic activity from the source locations in many cases is enhanced during ~150 - 750 ms following the TGFs, as also reported in Omar et al. (2014) and Smith et al. (2016). This enhanced activity is not present in a randomly-selected sample of flashes, suggesting it is a characteristic property of a significant fraction of flashes that start with a TGF.</p><p>The study is submitted to JGR Atmospheres.</p><p>Omar et al. (2014), <em>Characterizing the TGF-lightning relationship using ENTLN</em>, AGU Fall Meeting 2014, Abstract AE31A-3388.</p><p>Smith et al. (2016), doi:10.1002/2016JD025395.</p>

The Gamma-Flash program, funded by the Italian Space Agency (ASI) and led by the National Institute for Astrophysics (INAF), aims to study high-energy emissions related to thunderstorms, such as terrestrial gamma-ray flashes (TGFs) and gamma-ray glows. The program led to the development of two main detection systems: a ground-based system, installed at the “O. Vittori” Observatory on top of Mt. Cimone (Northern-Central Italy), and an airborne payload, installed on a Cessna Citation Mustang aircraft, for in-flight campaigns. The ground-based detection system consists of five γ-ray and three neutron detectors, and it has been collecting data from Jul 2022 to Oct 2023, overall experiencing 95 days of thunderstorm activity (36% of the total experiment lifetime). During this first observational survey, a gamma-ray glow of ∼ 1.5 min was revealed. The event light curve shows an abrupt interruption at the end, coinciding with the occurrence of a CG lightning discharge, that took place within 2 km from the detectors. The avionic payload consists of 6 γ-ray and 2 neutron detectors. The purpose of this second payload is to collect additional data by flying nearby convective systems, in order to reveal high-energy emissions directly from the sky. To date, two flights have been conducted: the first was an operational test flight on December 22, 2023, lasting 2 hours under clear sky conditions. The second flight, on June 4, 2024, lasted approximately 2.5 hours and took place during thunderstorm activity in Northern Italy. Additional flights are planned for the summer of 2024.

Unsupervised deep learning and semi-automatic data labeling in weed discrimination

Preliminary breakdown pulses (PBPs) occur in the initial phase of lightning. A realistic model for their description is employed to investigate relation between PBP peak currents and PBP electric field amplitudes and their relation to the return stroke (RS) peak currents. We demonstrate that the PBP peak currents can reach 200 kA and can be comparable or higher than the corresponding RS peak currents. For a typical PBP electric field waveform PBP peak currents are approximately proportional to the electric field amplitudes. We show that the PBP bipolar overshoot depends primarily on the characteristic time of the line conductivity increase. The magnitude of the charge centers is demonstrated to be very large in order to model the observed PBPs with amplitudes up to 32 V/m at 100 km. Such energetic current pulses might be capable to produce elves or terrestrial gamma ray flashes.

Fermi Gamma-ray Space Telescope has detected a diverse range of gamma-ray transients since its launch in 2008. Over the years, Fermi has accumulated an extensive public archive of transient events. Traditional classification methods for these events typically rely on fixed thresholds, localisation accuracy, and characteristic light-curve features. However, in the current era of time-critical, multi-wavelength, and multi-messenger astronomy, rapid and reliable classification is essential to enable timely follow-up and coordinated observations. In this work, we develop and present two deep learning-based classifiers that integrate convolutional and recurrent neural network architectures. Using multivariate time-series inputs derived from Fermi-GBM data, our models are trained to distinguish among four classes of gamma-ray transients: Gamma-Ray Bursts (GRBs), Terrestrial Gamma-ray Flashes (TGFs), Solar Flares (SFLAREs), and Soft Gamma Repeaters (SGRs). Furthermore, the models are designed to flag events that do not conform to any of these categories, providing a pathway for identifying potentially new or rare transient types.Training was conducted using a carefully curated subset of high-confidence Fermi events. The resulting models achieve an overall classification accuracy of 93%, and identify approximately 2.5% of the triggers as outliers of unknown origin. When applied to Fermi events with uncertain classifications, our models assign 60% of them to the TGF category with over 60% confidence. These results demonstrate that incorporating deep learning-based classification into onboard or automated data pipelines can significantly enhance transient identification, minimize misclassification, and improve the discovery potential of new phenomena in future high-energy astrophysics missions.

The evolution and improvements of deep learning are being used to tackle any research obstacles that could be converted into classification problems in all spheres of life. Each Deep convolutional neural network (DCNN) design's output is determined by the depth and value of the hyperparameters, which explains why so many of them have been proposed. These DCNN architectures must be created entirely from scratch, and they can only be used for the applications for which they were intended. Transfer learning may be used to modify these pre-trained networks so they are more appropriate for particular purposes. This article aims to evaluate the empirical performance of the applicability of pre-trained DCNN models to identify human face presentation threats (FPAD). Human FPAD is one of the most significant and crucial areas of research right now because of the introduction of ambient computing, which necessitates contact-free identification of persons with the help of their biometric traits. Six pre-trained DCNN models are taken into account for an experimental evaluation in human FPAD alias VGG19, VGG16, DensNet121, MobileNet, Xception, and InceptionV3. The investigation makes use of the NUAA and Replay-Attack benchmark FPAD datasets. Thepade's sorted block truncation coding (SBTC) 10-ary features are merged with deep learning features produced from the finest performing finetuned DCNNs to enhance the FPAD capabilities of analyzed machine learning (ML) classifiers. The integration of features of Thepade's SBTC 10-ary and DCNN has considerably increased the FPAD accuracy of ML classifiers with slightly more computations of feature extraction.

Early diagnosis and treatment is very essential in monitoring Brain tumor using MRI images. Convolutional Neural Networks (CNN) and Machine Learning (ML) classifiers have been widely used but there is not much work on how feature selection techniques would affect the performance of the CNN. Secondly, there is a need for investigation concerning small dataset adaptability and ML-CNN comparisons. To improve the classification accuracy, we integrate Univariate, Recursive Feature Elimination (RFE), Recursive Feature Elimination with Cross Validation (RFECV) with CNN in this study. Preprocessing, feature extraction & selection was carried out on the dataset consisting of 253 MRI images and they are classified using CNN and ML models (Logistic Regression, Decision Tree, Random Forest, Naïve Bayes). With the results 96%, CNN with Univariate Feature Selection performed better than ML classifiers, and other selection techniques. The results demonstrate that feature selection is necessary to get the best performance out of CNN models operating on small datasets. Future studies should be based on different deep learning architectures to improve classification and application in other datasets.

Machine learning has significantly impacted the analysis of biological images and is now an important part of many biological data analysis pipelines. A variety of biological and biomedical domain-related tasks is gaining benefit from image analysis and pattern recognition tools developed currently. Applications include diagnostic histopathology, environmental monitoring, synthetic biology, genomics, and proteomics. Particularly in the last decade, several deep learning and advanced computer vision methods such as convolutional neural networks (CNNs), typically trained in a supervised fashion, have started to be largely employed in biological image classification. Moreover, the advancement of automatic acquisition systems has been generating a massive amount of biological data, which requires to be analyzed by domain experts. However, the cost of manual annotation of such data has become a bottleneck, impairing the application of supervised machine learning algorithms. Biological images generally have an intrinsic high variability, whose identity is sometimes hard to assign and strongly dependent on the annotator’s expertise. In this context, a limited number of annotation-free (i.e., unsupervised) learning solutions have been proposed, typically based on hand-crafted features, specifically tailored for a certain biological domain. Nonetheless, a successful unsupervised learning approach must be accurate, and sufficiently robust to deal with different biological domains. This paper aims at providing a viable solution to these issues, proposing an unsupervised learning algorithm based on compressed deep features for image classification. We exploit features extracted from ImageNet pre-trained transformers and CNNs, further compressed with a customized β-Variational AutoEncoder (β-VAE), that we call reconstruction VAE (R-VAE). We test our algorithm on biological images coming from diverse domains characterized by high variability in shape and texture information and acquired with widely differing imaging platforms. Considered image datasets range from multi-cellular organisms (plankton, coral) to sub-cellular organelles (budding yeast vacuoles, human cells’ nuclei, etc.). Our results show that the compressed deep features extracted from different pre-trained vision models establish new unsupervised learning state-of-the-art performances for the investigated datasets.

Background: Early identification of probable post-traumatic stress disorder (PTSD) can lead to early intervention and treatment.Aims: This study aimed to evaluate supervised machine learning (ML) classifiers for the identification of probable PTSD in those who are serving, or have recently served in the United Kingdom (UK) Armed Forces.Methods: Supervised ML classification techniques were applied to a military cohort of 13,690 serving and ex-serving UK Armed Forces personnel to identify probable PTSD based on self-reported service exposures and a range of validated self-report measures. Data were collected between 2004 and 2009.Results: The predictive performance of supervised ML classifiers to detect cases of probable PTSD were encouraging when compared to a validated measure, demonstrating a capability of supervised ML to detect the cases of probable PTSD. It was possible to identify which variables contributed to the performance, including alcohol misuse, gender and deployment status. A satisfactory sensitivity was obtained across a range of supervised ML classifiers, but sensitivity was low, indicating a potential for false negative diagnoses.Conclusions: Detection of probable PTSD based on self-reported measurement data is feasible, may greatly reduce the burden on public health and improve operational efficiencies by enabling early intervention, before manifestation of symptoms.

This paper demonstrates exceptional performance of approximately 10.0 dB learning-based gain using the Deep Convolutional Neural Network (DCNN) for demodulation of a Rayleigh-faded wireless data signal. We simulate FSK demodulation over an AWGN Rayleigh fading channel with average signal to noise ratios (SNR) from 10 dB to 20 dB. The most recent and accurate classifier is the Deep Convolutional Neural Network (DCNN) which resulted in the lowest error bit probabilities between 0.00128 to 0.00019 for the range of SNRs. A comparative study has been applied between DCNN and other machine learning classifiers such as Support Vector Machine (SVM), Linear Discriminant Analysis (LDA), Multi-Layer Perceptron (MLP) which give bit error probabilities between 0.021 to 0.002, and Quadratic Discriminant Analysis (QDA) which gives bit error probabilities between 0.027 to 0.003. Frequency-shift keying (FSK) demodulation using matched filtering showed bit error probabilities between 0.025 to 0.0025. We also discuss the complexity issues with the DCNN regarding decoding rates and training set sizes. This work shows how much the DCNN would provide substantial benefit as the demodulator.

In recent years, COVID-19 has been regarded as the most dangerous pandemic for several countries. On various social media platforms, such as Twitter, Facebook, and Instagram, a variety of rumours, hypes, and news are published. This might have a detrimental impact on people’s life. As a result, social media platforms have always had a difficult time authenticating this fake information. Different machine learning (ML) and deep learning (DL) classifiers were used in this work to categorize the continuing impacts of tweets and forecast their after-effects. Support vector machine (SVM), random forest (RF), decision tree (DT), and k-nearest neighbor (KNN) were used for classification, while AdaBoost and convolutional neural network (CNN) were utilized for future effects. The tweets dataset from Kaggle was used to train the SVM, RF, KNN, and DT models, which were then assessed on multiple evaluation criteria such as accuracy, precision, recall, and F1-score, using a 70 : 30 ratio. The CNN and AdaBoost, on the other hand, have been taught to detect the mean square error, root mean square error, and mean absolute error. With 0.74 and 0.73 percent score out of 1, respectively, RF and SVM exhibit the best accuracy in impact when classifying the outcomes on the obtained dataset. In terms of a regression problem, CNN beat the ADA Regressor across the board.

As robots and machines become more reliable, developing tools that utilize their potential in manufacturing and beyond is an important step being addressed by many, including the LIMDA team at the University of Alberta. A common and effective means to improve artificial performance is through optical character recognition methods. Within the category of artificial intelligence under classification machine learning, research has focussed on the benefits of convolutional neural networks (CNN) and the improvement provided compared to its parent method, neural networks. Neural networks serious flaw comes from memorization and the lack of learning about what the images contain, while CNN's combat those issues. CNN’s are designed to connect information received by the network and begins to closely mimic how humans experience learns. Using the programming language Python and machine learning libraries such as Tensorflow and Keras, different versions of CNN’s were tested against datasets containing low-resolution images with handwritten characters. The first two CNN’s were trained against the MNIST database against digits 0 through 9. The results from these tests illustrated the benefits of elements like max-pooling and the addition of convolutional layers. Taking that knowledge a final CNN was written to prove the accuracy of the algorithm against alphabet characters. After training and testings were complete the network showed an average 99.34% accuracy and 2.23% to the loss function. Time-consuming training epochs that don’t wield higher or more impressive results could also be eliminated. These and similar CNN’s have proven to yield positive results and in future research could be implemented into the laboratory to improve safety. Continuing to develop this work will lead to better translators for language, solid text to digital text transformation, and aides for the visual and speech impaired.

Deep convolutional neural networks (DCNNs) for diagnosis of disease on chest radiographs (CXR) have been shown to be biased against males or females if the datasets used to train them have unbalanced sex representation. Prior work has suggested that DCNNs can predict sex on CXR, which could aid forensic evaluations, but also be a source of bias. To (1) evaluate the performance of DCNNs for predicting sex across different datasets and architectures and (2) evaluate visual biomarkers used by DCNNs to predict sex on CXRs. Chest radiographs were obtained from the Stanford CheXPert and NIH Chest XRay14 datasets which comprised of 224,316 and 112,120 CXRs, respectively. To control for dataset size and class imbalance, random undersampling was used to reduce each dataset to 97,560 images that were balanced for sex. Each dataset was randomly split into training (70%), validation (10%), and test (20%) sets. Four DCNN architectures pre-trained on ImageNet were used for transfer learning. DCNNs were externally validated using a test set from the opposing dataset. Performance was evaluated using area under the receiver operating characteristic curve (AUC). Class activation mapping (CAM) was used to generate heatmaps visualizing the regions contributing to the DCNN's prediction. On the internal test set, DCNNs achieved AUROCs ranging from 0.98 to 0.99. On external validation, the models reached peak cross-dataset performance of 0.94 for the VGG19-Stanford model and 0.95 for the InceptionV3-NIH model. Heatmaps highlighted similar regions of attention between model architectures and datasets, localizing to the mediastinal and upper rib regions, as well as to the lower chest/diaphragmatic regions. DCNNs trained on two large CXR datasets accurately predicted sex on internal and external test data with similar heatmap localizations across DCNN architectures and datasets. These findings support the notion that DCNNs can leverage imaging biomarkers to predict sex and potentially confound the accurate prediction of disease on CXRs and contribute to biased models. On the other hand, these DCNNs can be beneficial to emergency radiologists for forensic evaluations and identifying patient sex for patients whose identities are unknown, such as in acute trauma.

Due to the latest advancements in monitoring technologies, interest in the possibility of early-detection of quality issues in components has grown considerably in the manufacturing industry. However, implementation of such techniques has been limited outside of the research environment due to the more demanding scenarios posed by production environments. This paper proposes a method of assessing the health of a machining process and the machine tool itself by applying a range of machine learning (ML) techniques to sensor data. The aim of this work is not to provide complete diagnosis of a condition, but to provide a rapid indication that the machine tool or process has changed beyond acceptable limits; making for a more realistic solution for production environments. Prior research by the authors found good visibility of simulated failure modes in a number of machining operations and machine tool fingerprint routines, through the defined sensor suite. The current research set out to utilise this system, and streamline the test procedure to obtain a large dataset to test ML techniques upon. Various supervised and unsupervised ML techniques were implemented using a range of features extracted from the raw sensor signals, principal component analysis and continuous wavelet transform. The latter were classified using convolutional neural networks (CNN); both custom-made networks, and pre-trained networks through transfer learning. The detection and classification accuracies of the simulated failure modes across all classical ML and CNN techniques tested were promising, with all approaching 100% under certain conditions.

This research paper explores the hybrid models for Javanese character recognition using 15600 characters gathered from digital and handwritten sources. The hybrid model combines the merit of deep learning using convolutional neural networks (CNN) to involve feature extraction and a machine learning classifier using support vector machine (SVM). The dropout layer also manages overfitting problems and enhances training accuracy. For evaluation purposes, we also compared CNN models with three different architectures with multilayer perceptron (MLP) models with one and two hidden layer(s). In this research, we evaluated three variants of CNN architectures and the hybrid CNN-SVM models on both the accuracy of classification and training time. The experimental outcomes showed that the classification performances of all CNN models outperform the classification performances of both MLP models. The highest testing accuracy for basic CNN is 94.2% when using model 3 CNN. The increment of hidden layers to the MLP model just slightly enhances the accuracy. Furthermore, the hybrid model gained the highest accuracy result of 98.35% for classifying the testing data when combining model 3 CNN with the SVM classifier. We get that the hybrid CNN-SVM model can enhance the accuracy results in the Javanese characters recognition.