Improvements from incorporating machine learning algorithms into near real-time operational post-processing.

The cybersecurity concerns get increasingly intricate as the digital world progresses. In light of the increasing complexity of cyber threats, it is imperative to develop and implement advanced and flexible security strategies. Machine Learning (ML) has become a potent tool in strengthening cybersecurity, providing the capacity to scrutinise extensive information, recognise trends, and improve threat detection and defence methods. This paper examines the significance of ML in the field of cybersecurity, with a special emphasis on the identification of threats and the implementation of protective measures. By incorporating ML algorithms into cybersecurity frameworks, organisations may automate decision-making processes, facilitating prompt responses to ever-changing threats. The initial segment explores the terrain of cyber threats, highlighting the necessity for dynamic and aggressive security methods. Conventional solutions that rely on signatures are frequently inadequate when it comes to handling sophisticated, shape-shifting attacks. ML algorithms, in contrast, have exceptional proficiency in identifying nuanced patterns and irregularities within extensive datasets, therefore offering a more efficient method of detecting potential threats. The second section delves into several ML methodologies utilised in cybersecurity, including supervised and unsupervised learning, deep learning, and reinforcement learning. Every approach is assessed based on its suitability for threat detection, demonstrating its advantages and constraints. Furthermore, the relevance of feature engineering and data pretreatment in improving machine learning models for cybersecurity applications. The versatility of ML algorithms allows them to grow with emerging threats, making them a useful tool in the ever-changing arena of cyber warfare. The final segment focuses on real-world applications of machine learning in cybersecurity, presenting successful use cases across sectors. From anomaly detection to behavior analysis, ML algorithms contribute to the discovery of dangerous activity, lowering false positives and strengthening the overall security posture. Lastly, the paper covers the obstacles and ethical issues related to the adoption of ML in cybersecurity. Issues like as adversarial assaults, skewed datasets, and the interpretability of ML models are examined, highlighting the necessity for a holistic strategy that integrates modern technology with ethical considerations. The fusion of human expertise and machine intelligence offers a formidable defense against evolving cyber threats, paving the way for a more resilient and secure digital future.

The Software-Defined Networking technology promises to enhance network performance and cost reduction for service providers by providing scalability, flexibility, and programmability through the separation of the control plane from the data plane. However, the separation between the control plane and data plane in the implementation of SDN presents scalability issues, as the controller has limited computational resources. To address the SDN scalability issues identified, we create a scalable hybrid switching solution using machine learning algorithms. We propose an SDN OpenFlow model switch which collaborates with the traditional switch to represent a scalable framework of Hybrid Routing with Reinforcement Learning (sHRRL). We implement a reinforcement algorithm to randomly explore new routes and discover the most optimal path through the Q-learning algorithm. This primitive and model-free form of reinforcement learning utilizes the Markov Decision Process and Bellman’s equation to reiteratively update Q-values in the Q-table for every transition in the network environment state until Q-function has converged to the best Q-values. The proposed hybrid switching model is benchmarked against the standard SDN OpenFlow switch in terms of network performance metrics, including throughput, packet exchange transmission rates, CPU load, and delay. When statistically comparing simulation results, it is evident that the proposed switching model, incorporating machine learning algorithms, can effectively tackle scalability challenges in the design of SDN controller networks, especially in Data Centre environments where rapid switching speeds are crucial.

Implementing machine learning for the identification and classification of compound and mixtures in portable Raman instruments

Incorporation of high accuracy surface modeling into machine learning to improve soil organic matter mapping

This paper aims to present an integrated approach for well forecasting that combines advanced Machine Learning (ML) techniques to provide accurate oil, gas and water predictions for existing and new wells in oil reservoirs. Well forecasting is a critical task in the oil and gas industry for optimizing production and maximizing recovery from reservoirs. Traditionally, decline curve analysis has been widely used for well forecasting, but it has limitations in complex reservoirs or wells with uncertain decline trends. The proposed approach overcomes these limitations by utilizing advanced Machine Learning techniques that leverage complex real data for improved production forecasting. By incorporating ML algorithms, the proposed approach can provide more accurate and reliable predictions, particularly in challenging reservoir conditions where traditional methods may fall short. The proposed approach for well forecasting involves several key steps. Firstly, historical data of wells, including production, choke, well head flowing pressure, producing days, injection data, well connection properties, and 3D properties, are collected and preprocessed. Next, 2D properties, such as well location, spatial coordinates, and spatial attributes, are extracted from the 3D properties of the reservoir using geospatial algorithms. Then, for existing wells an ensemble of ML supervised algorithm is trained well by well with enough historical data, and nearby wells are used to train non mature wells. For future new wells, historical data plus geological 2D properties is used to create a model per reservoir. This involves collecting and preprocessing data from nearby wells in the vicinity of the new drilling location to obtain relevant properties. Uncertainty intervals are calculated using quantiles to quantify the uncertainty in the predictions. The integrated well forecasting approach presented in this paper provides a comprehensive solution for predicting well production in oil reservoirs. We used 2 different approaches to predict oil, gas and water production depending on the availability of historical data. Our approach allowed accurate predictions in complex reservoirs where wells have no clear decline trends. The use of uncertainty intervals provides a measure of prediction uncertainty, allowing for informed decision-making.

The Virtual Seismologist method for earthquake early warning uses a Bayesian approach to find the most probable magnitude and location estimates given the incoming ground motions envelopes from a rupturing earthquake. Ground motion ratios and ground motion envelope attenuation relationships are used to estimate magnitude and epicentral location as early as 3 seconds after the initial P wave detection. The use of prior information distinguishes this method from other proposed methods for seismic early warning. The state of health of the seismic network, previously observed seismicity, fault locations, and the Gutenberg-Richter relationship are the types of prior information useful in resolving trade-offs in the initial source estimates which are unresolved by the limited data. Short-term earthquake forecasts are ideal priors for seismic early warning. Having a high density of stations with real-time telemetry reduces the complexity involved in finding the most probable source estimates and communicating these estimates to early warning subscribers. The benefits of prior information are most evident for regions with low station density. Most early warning studies are focused exclusively on either the source estimation problem, or how subscribers use the warning information. The inclusion of prior information ultimately requires a level of coordination and communication between the network broadcasting the early warning information and the subscribers that is not consistent with this divide. The need for a more integrated approach to seismic early warning which considers the source estimation and user response as interacting and interrelated parts of a single problem is discussed. A parameterization that decomposes observed ground motion envelopes into Pwave, S-wave, and ambient noise envelopes is developed and applied to a large suite of observed ground motion envelopes recorded within 200 km of 2 ! M ! 7.3 Southern California earthquakes. Separate attenuation relationships are developed to describe vi the magnitude, distance, and site dependence of various channels of P- and S-wave envelopes. The P-wave relationships allow the early warning source estimates to be obtained from observed P-wave amplitudes. Aside from early warning applications, these envelope attenuation relationships are used to investigate the average properties of ground motions recorded by the Southern California Seismic Network. Stationspecific amplification factors for 150 Southern California Seismic Network stations were obtained for horizontal and vertical acceleration, velocity, and displacement amplitudes, and are included (Excel format) as external multimedia objects.

The Pasadena Office of the U.S. Geological Survey, together with the California Institute of Technology Seismology Laboratory (Caltech Seismo Lab), operates a network of more than 300 remote seismometers in southern California called the Southern California Seismic Network (SCSN). Signals from these sites are telemetered to the central processing site at the Caltech Seismology Laboratory in Pasadena. These signals are continuously monitored by computers that detect and record thousands of earthquakes each year. Phase arrival times for these events are picked by analysts and archived along with digital seismograms. Data acquisition, processing, and archiving is achieved using the CUSP system (Dollar, 1989). These data are used to compile the SCSN Catalog of Earthquakes , a list beginning in 1932 that currently contains more than 282,000 events. This data set is critical to the evaluation of earthquake hazards in California and to the advancement of geoscience as a whole. This article contains...

Machine-learning algorithms are widely applied to facilitate human tasks. For instance, in seismology, they can help deliver high-resolution seismic catalogs including very small magnitude events that usually remain undetected by human analysts and by standard monitoring procedures. The new frontier of modern seismology is to exploit deep neural networks (DNN) to automatically detect P- and S-wave arrival times to obtain good-quality seismic event locations in terms of hypocentral errors. Increasing the number of events and seismic phases is essential to build complete earthquake catalogs to be used in seismological analyses (such as seismic hazard estimation, seismic tomography, fault zone structure determination, rupture mechanism study, etc.). Machine Learning methods are being integrated into large Research Infrastructures (RIs), like the European Plate Observing System (EPOS ERIC), which brings together 10 different scientific domains in Solid Earth Sciences. In this contribution, we present results from a specific Sponsored Research Activity promoted by the RI EPOS and dedicated to ML-driven methods for phase picking in seismic time series.To ensure the correct recognition of seismic waves, neural networks trained on large and representative training datasets are essential. Here, we investigate the influence of the training dataset on the DNN PhaseNet performance, applying the method to three case studies. In the first case, the DNN trained with the Italian seismicity dataset (INSTANCE) is used to build a catalogue on the Fucino basin (Central Italy) study area; in the second case, we use the AQ2009 training dataset (based on the L’Aquila 2009 aftershocks) to analyse the 2016-2017 Amatrice-Visso-Norcia seismic sequence; and finally, the CREW training dataset (that contains P- and S-waves reflected on the mantle Earth and recorded at large epicentral distance) to detect P- and S- waves of teleseismic (regional) data acquired by the Adria Array project network. The use of different training datasets greatly improves the performance of the neural network in recognizing P and S phases, reducing the number of false positives and providing more accurate and precise P and S-phase arrival times. 

This article describes the programs included in the COMPLOC computer program package that are designed to apply the source-specific station term (SSST) method to solve for local earthquake locations using P- and S- wave phase data. These programs can greatly improve the relative location accuracy of nearby events by applying empirical corrections for the biasing effects of three-dimensional velocity structure. They have been tested on data from both the Southern California Seismic Network (SCSN) and Northern California Seismic Network (NCSN) The SSST method (Richards-Dinger and Shearer 2000; Lin and Shearer 2005) works by assigning each station a travel-time correction that varies as a function of source position. This approach provides relative location accuracy comparable to master event or hypocentroidal decomposition (Jordan and Sverdrup 1981) methods within compact event clusters, but is applicable to distributed seismicity. It has some similarities to the double-difference algorithm (Waldhauser and Ellsworth 2000, 2002) and can be shown to give comparable results in tests on synthetic data (Lin and Shearer 2005). However, the SSST approach has computational advantages for big data sets because the location and station term parts of the computation are separate, so that large matrix inversions are not necessary. In addition, our implementation of SSST provides the option to use L1-norm misfit measures, which are more robust than least squares in the case of occasional timing errors or bad phase picks. We implement the SSST approach by selecting nearby events located within a sphere of specified radius r max around the target event. The station term for the target event is then computed as the median (or mean) residual of these events. Different results will be obtained depending upon the size of the cutoff distance r max. If r max is set to a large enough distance, then the SSST …

Predicting intraurban PM2.5 concentrations using enhanced machine learning approaches and incorporating human activity patterns

This study presents a comparative analysis of the classification performance of facial and emotion recognition systems using Machine Learning (ML) and Deep Learning (DL) algorithms. The primary objective of this work is to evaluate the applicability of emotion recognition in fields such as psychotherapy and crime analysis, using the FER-2013 dataset.The study was conducted with ML algorithms such as Support Vector Machines (SVM), Random Forest (RF), K-Nearest Neighbor (KNN), Decision Trees, and Gradient Boosting, as well as the DL algorithm, Convolutional Neural Network (CNN). Supported by the application of feature selection, preprocessing, and feature extraction techniques, the model’s performance was measured using standard metrics such as accuracy, precision, F1-score, and AUC-ROC.The experimental results showed that the highest classification accuracy (46.61%) was achieved with the CNN model. While the ML models generally offered lower accuracy, they provided advantages in terms of computational efficiency in specific scenarios.This study aims to contribute to the literature by providing a comparative analysis of ML and DL algorithms and by highlighting the effect of data preprocessing on performance. The findings set targets for future work, such as real-time system integration and hybrid model development.

In real-time image processing, the demand for efficient solutions has surged with the proliferation of applications spanning from autonomous vehicles to medical diagnostics. This study addresses the imperative need for accelerated machine learning algorithms to enhance the processing speed of image-related tasks. The research focuses on leveraging Field-Programmable Gate Arrays (FPGAs) to implement hardware acceleration, exploiting their parallel computing capabilities. The advent of machine learning in image processing has revolutionized various industries, yet real-time applications encounter computational bottlenecks. This research delves into hardware acceleration using FPGAs to overcome these constraints, offering a novel approach to expedite machine learning algorithms. Traditional software implementations of machine learning algorithms often fall short in meeting real-time processing requirements. This research aims to bridge this gap by exploring FPGA-based hardware acceleration, addressing the performance limitations hindering the seamless integration of machine learning into real-time image processing systems. While existing literature acknowledges the potential of FPGA-based acceleration, a comprehensive exploration of its application for real-time image processing is lacking. This research fills the void by presenting a detailed method and empirical results, contributing to the limited body of knowledge on FPGA-accelerated machine learning in the of image processing. The study employs a systematic approach, integrating machine learning algorithms onto FPGAs through hardware description languages. The implementation is optimized to exploit parallelism inherent in FPGAs, resulting in a tailored hardware solution for real-time image processing. Comparative analyses against software implementations provide insights into the performance gains achieved. The experimental results demonstrate a significant enhancement in processing speed, validating the efficacy of FPGA-based hardware acceleration for machine learning algorithms in real-time image processing applications.

Breast cancer remains a leading cause of cancer-related deaths among women, emphasizing the critical need for early detection and monitoring techniques. Conventional imaging modalities such as mammography, MRI, and ultrasound have face sensitivity, specificity, cost, and patient comfort limitations. This study introduces a handheld Near-Infrared Diffuse Optical Tomography (NIR DOT) probe for breast cancer imaging. The NIRscan probe utilizes multi-wavelength light-emitting diodes (LEDs) and a linear charge-coupled device (CCD) sensor to acquire real-time optical data, reconstructing cross-sectional images of breast tissue based on scattering and absorption coefficients. With wavelengths optimized for the differential optical properties of tissue components, the probe enables functional imaging, distinguishing between healthy and malignant tissues. Clinical evaluations have demonstrated its potential for precise tumor localization and monitoring therapeutic responses, achieving a sensitivity of 94.7% and specificity of 84.2%. By incorporating machine learning algorithms and a modified diffusion equation (MDE), the system enhances the accuracy and speed of image reconstruction, supporting rapid, non-invasive diagnostics. This development represents a significant step forward in portable, cost-effective solutions for breast cancer detection, with potential applications in low-resource settings and diverse clinical environments.

The fusion of unmanned aerial system (UAS) and satellite imagery has emerged as a pivotal strategy in advancing precision agriculture. This review explores the significance of integrating high-resolution UAS and satellite imagery via pixel-based, feature-based, and decision-based fusion methods. The study investigates optimization techniques, spectral synergy, temporal strategies, and challenges in data fusion, presenting transformative insights such as enhanced biomass estimation through UAS-satellite synergy, improved nitrogen stress detection in maize, and refined crop type mapping using multi-temporal fusion. The combined spectral information from UAS and satellite sources confirms instrumental in crop monitoring and biomass estimation. Temporal optimization strategies consider factors such as crop phenology, spatial resolution, and budget constraints, offering effective and continuous monitoring solutions. The review systematically addresses challenges in spatial and temporal resolutions, radiometric calibration, data synchronization, and processing techniques, providing practical solutions. Integrated UAS and satellite data impact precision agriculture, contributing to improved resolution, monitoring capabilities, resource allocation, and crop performance evaluation. A comparative analysis underscores the superiority of combined data, particularly for specific crops and scenarios. Researchers exhibit a preference for pixel-based fusion methods, aligning fusion goals with specific needs. The findings contribute to the evolving landscape of precision agriculture, suggesting avenues for future research and reinforcing the field’s dynamism and relevance. Future works should delve into advanced fusion methodologies, incorporating machine learning algorithms, and conduct cross-crop application studies to broaden applicability and tailor insights for specific crops.

The rapid urbanization and technological advancement have catalyzed the emergence of smart cities as a transformative paradigm for sustainable urban development. This paper presents a comprehensive framework for modeling smart cities through the systematic integration of Internet of Things (IoT), big data, and analytics technologies. We propose a multi-layered architectural model that addresses the technical, operational, and governance challenges inherent in smart city implementations. The research examines how IoT sensors and devices generate massive volumes of heterogeneous data, which are subsequently processed through big data platforms to extract actionable insights via advanced analytics techniques. Our framework encompasses data acquisition, storage, processing, and visualization layers, while incorporating machine learning algorithms and real-time analytics for intelligent decision-making. Through case studies of various smart city domains including transportation, energy management, public safety, and healthcare, we demonstrate the practical applicability of our integrated model. The paper also addresses critical challenges such as data privacy, security, interoperability, and scalability that must be overcome for successful smart city deployment. Our findings reveal that effective integration of these three technological pillars enables cities to optimize resource allocation, enhance service delivery, improve quality of life for citizens, and achieve sustainability goals. The proposed model provides urban planners, policymakers, and technology implementers with a structured approach to design and deploy smart city solutions that are both technologically robust and contextually relevant.