Search

In the HEP community the prediction of Data Popularity is a topic that has been approached for many years. Nonetheless, while facing increasing data storage challenges, especially in the upcoming HL-LHC era, there is still the need for better predictive models to answer the questions of whether particular data should be kept, replicated, or deleted. Caches have proven to be a convenient technique for partially automating storage management, potentially eliminating some of these questions. On the one hand, one can benefit even from simple cache eviction policies like LRU, on the other hand, we show that incorporation of knowledge about future access patterns has the potential to greatly improve cache performance. In this paper, we study data popularity on the file level, where the special relation between files belonging to the same dataset could be used in addition to the standard attributes. We turn to Machine Learning algorithms, such as Random Forest, which is well suited to work with Big Data: it can be parallelized, is more lightweight and easier to interpret than Deep Neural Networks. Finally, we compare the results with standard cache eviction algorithms and the theoretical optimum. [ABSTRACT FROM AUTHOR]

Our study introduces a new machine learning algorithm for estimating 3D cumulative radial profiles of total and gas mass in galaxy clusters from thermal Sunyaev-Zel'dovich (SZ) effect maps. We generate mock images from 2522 simulated clusters, employing an autoencoder and random forest in our approach. Notably, our model makes no prior assumptions about hydrostatic equilibrium. Our results indicate that the model successfully reconstructs unbiased total and gas mass profiles, with a scatter of approximately 10%. We analyse clusters in various dynamical states and mass ranges, finding that our method's accuracy and precision are consistent. We verify the capabilities of our model by comparing it with the hydrostatic equilibrium technique, showing that it accurately recovers total mass profiles without any bias. [ABSTRACT FROM AUTHOR]

Identifying changes in the Earth's phenomena is vital for understanding and mitigating the impacts of environmental issues. Monitoring the Earth's surface phenomena can be carried out effectively using satellite images acquired at different times. In addition to spectral features, spatial features play a significant role in detecting precise changes. However, classical change detection (CD) methods rarely consider spatial information and fail to account for scale variations within images. The present study introduces a novel deep learning-based CD method that hierarchically extracts spatial–spectral features at various scales to address these issues. The proposed deep neural network generates a binary change map by employing a multi-scale approach that integrates the information of patches of varied sizes at the decision level. We conducted experiments using Landsat-8 images from Sahand City, East Azarbaijan, Iran, because of their remarkable capacity to represent the Earth's surface details. Tabriz's population growth has led to rapid development in Sahand City to accommodate citizens. Studying these changes can offer valuable insights into urban planning. The performance of the proposed deep model is evaluated in comparison to two classical methods, the change vector analysis (CVA) method and a random forest (RF) algorithm. Based on the change detection results, the proposed deep learning network demonstrates a significant improvement in the kappa coefficient (KC) compared to the RF and CVA methods, with increases of approximately 11.86% and 29.36%, respectively. Furthermore, in terms of overall accuracy (O.A.), the proposed network outperforms both the RF and CVA methods by approximately 17.08% and 29.16%, respectively. The proposed multi-scale deep network performs better at detecting changes across all metrics. As a result, the CVA method fails to identify changes with sufficient accuracy. [ABSTRACT FROM AUTHOR]

This research evaluates the capability of Sentinel-2 satellite imagery for mapping Cymodocea nodosa meadows in El Médano (Tenerife, Canary Islands, Spain). A Level-1C image from 27 October 2022 was used. Atmospheric correction was addressed using the Sen2Cor tool, while Lyzenga's method was employed to account for the water column effect. Three supervised classifications were performed using Random Forest, K-Nearest Neighbors (KNN) and KDTree-KNN algorithms. These classifications were complemented by an unsupervised classification and in situ data. Additionally, the amount of blue carbon sequestered by the C. nodosa in the study area was also estimated. Among the classifiers, the Random Forest algorithm produced the highest F1 scores, ranging from 0.96 to 0.99. The results revealed an average area of 237 ± 5 ha occupied by C. nodosa in the study region, translating to an average sequestration of 111,000 ± 2000 Mg CO2. Notably, the seagrass meadows in this study area have the potential to offset the CO2 emissions produced by the industrial combustion plant sector throughout the Canary Islands. This research represents a significant step forward in the protection and understanding of these invaluable ecosystems. It effectively underlines the potential of Sentinel-2 satellite data to map seagrass meadows and highlights their crucial role in achieving net zero carbon emissions on our planet. [ABSTRACT FROM AUTHOR]

This paper presents a meticulous exploration of advanced machine learning techniques for precise forest type classification using multi-temporal remote sensing data within a woodland environment. The study comprehensively evaluates a diverse range of models, ranging from advanced (ensemble) machine learning (ML) methods to several finely tuned support vector machine (SVM) variants, with a specific focus on Bayesian-optimized SVM with a radial basis function (RBF) kernel. Our findings highlight the robust performance of the Bayesian-optimized SVM, achieving a high accuracy of up to 94.27% and average precision and recall of 94.46% and 94.27%, respectively. Notably, this accuracy aligns with the levels attained by acclaimed ensemble techniques such as random forest and CatBoost while also surpassing those of XGBoost and LightGBM. These results highlight the potential of these methodologies to significantly enhance forest type mapping accuracy compared to traditional (linear) SVM and black-box neural networks. This, in turn, can enable the reliable identification and quantification of key services, including carbon storage and erosion protection, intrinsic to the forest ecosystem. The findings of our comparative study emphasize the profound impact of employing and fine-tuning ML approaches in the realm of remote sensing-based environmental analysis. [ABSTRACT FROM AUTHOR]

Ensuring food security in precision agriculture demands early prediction of corn yield in the USA at international, regional, and local levels. Accurate corn yield estimation can play a crucial role in averting famine by offering insights into food availability during the growing season. To address this, we propose a Concatenate-based 2D-CNN-BILSTM model that integrates Sentinel-1, Sentinel-2, and Soil GRIDS (global gridded soil information) data for corn yield estimation in Iowa State from 2018 to 2021. This approach utilizes Sentinel-2 features, including spectral bands (Blue, Green, Red, Red Edge 1/2/3, NIR, n-NIR, and SWIR 1/2), and vegetation indices (NDVI, LSWI, DVI, RVI, WDRVI, SAVI, VARIGREEN, and GNDVI), alongside Sentinel 1 features (VV, VH, difference VV, and VH, and RVI), and soil data (Silt, Clay, Sand, CEC, and pH) as initial inputs. To extract high-level features from this data each month, a dedicated 2D-CNN was designed. This 2D-CNN concatenates high-level features from the previous month with low-level features of the subsequent month, serving as input features for the model. Additionally, to incorporate single-time soil data features, another 2D-CNN was implemented. Finally, high-level features from soil, Sentinel-1, and Sentinel-2 data were concatenated and fed into a BILSTM layer for accurate corn yield prediction. Comparative analysis against random forest (RF), Concatenate-based 2D-CNN, and 2D-CNN models, using metrics like RMSE, MAE, MAPE, and the Index of Agreement, revealed the superiority of our model. It achieved an Index of Agreement of 84.67% with an RMSE of 0.698 t/ha. The Concatenate-based 2D-CNN model also performed well with an RMSE of 0.799 t/ha and an Index of Agreement of 72.71%. The 2D-CNN model followed closely with an RMSE of 0.834 t/ha and an Index of Agreement of 69.90%. In contrast, the RF model lagged with an RMSE of 1.073 t/ha and an Index of Agreement of 69.60%. Integration of Sentinel 1–2 and Soil-GRIDs data with the Concatenate-based 2D-CNN-BILSTM model significantly improved accuracy. Combining soil data with Sentinel 1–2 features reduced the RMSE by 16 kg and increased the Index of Agreement by 2.59%. This study highlighted the potential of advanced machine learning (ML)/deep learning (DL) models in achieving precise and reliable predictions, which could support sustainable agricultural practices and food-security initiatives. [ABSTRACT FROM AUTHOR]

The first talk at the session Random trees and random forests "Journée MAS" (27/08/2021) was presented by I. Kortchemski. After a general up-to-date introduction to local and scaling limits of Bienaymé trees (which are discrete branching trees), he presented new results on precise behavior of the largest out-degree of large branching trees when the offspring distribution μ is subcritical with μ(n) of order n−β for large n and β > 2 or critical with μ(n) of order n−2. In the next talk, M. Nassif gave asymptotics of additive functionals of large Bienayme trees in the global regime, which can be understood using scaling limits. Looking at Cayley trees with fixed size, A. Contat established a surprising identity for randomly built independent sets. Eventually J.-J. Duchamps presented some results on the distribution of the discrete Moran forest, a random graph arising in a classical population model at equilibrium. [ABSTRACT FROM AUTHOR]

The application of machine learning (ML) algorithms in large datasets in the field of meteorology is at the forefront of research. In this context, the use of satellite data to estimate the amount of rainfall is an important field of research, with operational applications. It is important to accurately predict the amount of rainfall (or rain rate) in a particular area for the proper taking of life and property protection measures. The present work intends to deepen the analysis of meteorological data with ML techniques to improve our capacity in short-range forecasting of rainfall. To this end, relationships between thermodynamic parameters derived by satellite measurements and recorded rainfall by in situ gauges, along with outputs from a numerical atmospheric model are analyzed. The main purpose of the work is to find the best relationships between the atmospheric conditions and the formation of clouds that lead to production of rainfall and build a ML model for nowcasting of rainfall. Several ML methods are used, i.e., Auto Regression, Ensemble Machine Learning, and Deep Learning, and their results are compared in order to find the best fit model. [ABSTRACT FROM AUTHOR]

With the exponential growth of digital media, readers face a daunting task of sifting through vast amounts of information to identify important news. This problem is especially critical for media professionals, journalists, and news agencies who need to quickly filter news articles to identify relevant and significant stories. Machine learning models offer a promising solution by automatically classifying news articles based on their significance. In this paper, we propose novel machine learning models for news significance detection, leveraging state-of-the-art deep learning architectures and a dataset of news articles. We evaluate our models using a variety of performance metrics and demonstrate their effectiveness compared to existing methods. Our proposed approach has the potential to significantly improve the efficiency and accuracy of news selection, benefiting both media professionals and readers alike. Furthermore, it can be beneficial to forecast the popularity of fake news and prevent its dissemination in society. Approximately, 2800 Azerbaijani news articles have been collected from telegram and labeled as popular or unpopular according to statistical calculation results. For news popularity detection, application of SVM, Random Forest and Neural network models and their results have been discussed in this paper. [ABSTRACT FROM AUTHOR]

In rural areas, water treatment plants use rudimentary techniques to evaluate turbidity. However, the incorrect measurement of turbidity can result in poor water quality and, as a result, health issues for its users because it is a crucial indicator to determine the application of adequate treatment to the water. Aquarisc was a project financed with royalties that sought to strengthen the mechanisms and tools for decision-making of the authorities and territorial institutions related to water supply for human consumption. This project installed sensors to assess turbidity in some plants in rural areas of the department of Cauca, Colombia. However, when the project ended, these sensors were removed. Therefore, it became necessary to create machine learning models to predict turbidity values without sensors, considering only pH, temperature, vapor pressure, and precipitation data captured manually by plant operators. In this study, the Linear Regression, Random Forest Regressor, k-Neighbors Regressor, and Extra Trees Regressor algorithms were trained with data provided by the Aquarisc project and the Institute of Hydrology, Meteorology, and Environment Studies of Colombia (IDEAM). As a result, we selected the Random Forest Regressor since it had the best RMSE among all the models and was also the one that best matched the situation of the studied treatment plants. Furthermore, this model did not consider outliers, resulting in an RMSE of 20.98 and 3.49 for the training and test dataset, respectively. Finally, we determined that this algorithm was able to estimate the water's turbidity acceptably and supports the operators in making decisions for the application of adequate treatment to drinking water. [ABSTRACT FROM AUTHOR]

Machine learning (ML)-based data-driven modelling is an efficient approach for good estimates of flow and maximum discharge at certain points within a basin. This paper is mainly aimed at evaluating the predictive capability of ML algorithms for daily streamflow modelling in the Besós River Basin (Spain), based on open source flow discharge and rainfall historical time series. In this sense, two modelling scenarios, without and with consideration of the antecedent hydrologic conditions, were evaluated, and three ML algorithms—support vector machines, random forest (RF) and gradient boosting (GB)—were compared to multiple linear regression (MLR), and were implemented. The prediction results revealed that the SVR model outperformed the other suggested models. Additionally, it was deduced that taking into account preceding hydrologic conditions clearly improves prediction results. [ABSTRACT FROM AUTHOR]

The emergence of Machine learning (ML) algorithms has shown competency in a variety of fields and are growing in popularity in their application to geospatial science issues. Most recently, and notably, ML algorithms have been applied to flood susceptibility (FS) mapping. Leveraging high-power computing systems and existing ML algorithms with national datasets of Canada, this project has explored methods to create a national FS layer across a geographically large and diverse country with limited training data. First, approaches were considered on how to generate a map of FS for Canada at two different levels, (i) national, which combined all training data into one model, and (ii) regional, where multiple models were created, based on regional similarities, and the results were mosaicked to generate a FS map. The second experiment explored the predictive capability of several ML algorithms across the geographically large and diverse landscape. Results indicate that the national approach provides a better prediction of FS, with 95.7% of the test points, 91.5% of the pixels in the training sites, and 89.6% of the pixels across the country correctly predicted as flooded, compared to 65.5%, 80.6% and 75.6%, respectively, in the regional approach. ML models applied across the country found that support vector machine (svmRadial) and Neural Network (nnet) performed poorly in areas away from the training sites, while random forest (parRF) and Multivariate Adaptive Regression Spline (earth) performed better. A national ensemble model was ultimately selected as this blend of models compensated for the biases found in the individual models in geographic areas far removed from training sites. [ABSTRACT FROM AUTHOR]

Predicting human response to complex sound is a nontrivial task. Besides large differences among subjects and practically infinite types of stimuli, human response to sound is typically quantified by a few parameters having nonlinear behavior. Still, such predictions are valuable for the assessment of the perception of sound, which is a critical step toward the development of systems that offer improved human comfort, productivity and wellness. The overall objective of this work is to learn about human perception of sound through the use of machine learning algorithms. Learning algorithms are ideal for modeling the complex behavior of subjective parameters and identifying new trends in behavior using knowledge accumulated from different experiments. This work compares the performance of four learning algorithms (linear regression (LR), support vector machines (SVM), decision trees (DTs), and bagged DTs/random forests (BDTRF)) used to construct models capable of predicting annoyance due to complex sound. Construction of models relied on annoyance responses of 38 subjects to 103 sounds described by five predictors (loudness, roughness, sharpness, total tone prominence, and fluctuation strength). Comparison of these models in terms of prediction accuracy, model interpretability, simplicity and versatility indicated that BDTRF is the best algorithm for this task. [ABSTRACT FROM AUTHOR]

Breiman's (2001) random forests are a very popular class of learning algorithms often able to produce good predictions even in high-dimensional frameworks, with no need to accurately tune its inner parameters. Unfortunately, there are no theoretical findings to support the default values used for these parameters in Breiman's algorithm. The aim of this paper is therefore to present recent theoretical results providing some insights on the role and the tuning of these parameters. [ABSTRACT FROM AUTHOR]

In order to perform a supervised classification of variable stars, we propose and evaluate a set of six features extracted from the magnitude density of the light curves. They are used to train automatic classification systems using state-of-the-art classifiers implemented in the R statistical computing environment. We find that random forests is the most successful method to select variables. [ABSTRACT FROM AUTHOR]