Examinando por Autor "Nicolis, Orietta"
Mostrando 1 - 20 de 20
Resultados por página
Opciones de ordenación
Ítem Analyzing the Impact of COVID-19 on Economic Sustainability: A Clustering Approach(MDPI, 2024-02-04) Nicolis, Orietta; Maidana, Jean Paul; Contreras, Fabian; Leal, DaniloThis work presents a comprehensive analysis of the economic impact of the COVID-19 pandemic, with a focus on OECD countries and the Chilean case. Utilizing a clustering approach, the research aims to investigate how countries can be categorized based on their pandemic mitigation strategies, economic responses, and infection rates. The methodology incorporates k-means and hierarchical clustering techniques, along with dynamic time warping, to account for the temporal variations in the pandemic’s progression across different nations. The study integrates the GDP into the analysis, thereby offering a perspective on the relationship between this economic indicator and health measures. Special attention is given to the case of Chile, thus providing a detailed examination of its economic and financial indicators during the pandemic. In particular, the work addresses the following main research questions: How can the OECD countries be clustered according to some health and economical indicators? What are the impacts of mitigation measures and the pension fund withdrawals on the Chilean economy? The study identifies significant differences (p-value < 0.05%) in the GDPs and infection rates between the two identified clusters that are influenced by government measures, particularly in the banking sector (55% and 60% in clusters 1 and 2, respectively). In Chile, a rebound in the IMACEC index is noted after increased liquidity, especially following partial pension fund withdrawals, thereby aligning with discrepancies between model forecasts and actual data. This study provides important insights for evidence-based public policies, thus aiding decision makers in mitigating the socioeconomic impact of global health crises and offering strategic advice for a sustainable economy.Ítem Detección y localización de terremoto mediante análisis de datos de Twitter(Universidad Andrés Bello, 2021) Castillo Ducaud, Felipe A.; Inostroza Negrete, Sergio E; Nicolis, Orietta; Facultad de IngenieríaPara muchos países, es importante estar preparados en caso de que ocurra un sismo y de esta forma reducir el daño producido. Países como Japón, China, Estados Unidos y Chile, entre otros, invierten gran cantidad de recursos en el estudio de terremotos y hay una constante necesidad de métodos novedosos para recolectar información rápidamente durante los momentos de crisis. Actualmente la principal fuente de información consiste en redes de sensores sísmicos de alta sensibilidad, que son capaces de detectar movimientos imperceptibles por las personas. Mediante el uso de sensores se puede medir la magnitud de un sismo. En Chile esta tarea recae en voluntarios de las oficinas de emergencia nacionales (ONEMI), quienes envían un reporte con la intensidad que percibieron en el área geográfica en la que se encuentran, y en otros países existen iniciativas que, por medio del crowdsourcing, determinan si el sismo fue perceptible. Esta información es valiosa porque permite a sismólogos estimar de mejor manera la fuerza con que un sismo fue percibido, el daño causado y el área de impacto, así como también establecer relaciones entre el tamaño de un sismo y los factores antes mencionados. Cabe destacar que, sin importar si un sismo es de alta o baja magnitud, su estudio y caracterización, permite a los expertos obtener conocimiento sobre la actividad sísmica de ciertas regiones y construir catálogos más completos sobre sismos en el mundo. Las oficinas de emergencia nacionales se esfuerzan en determinar la intensidad de un sismo en todos los lugares donde este fue percibido, incluso si se trata de un sismo pequeño. Este conocimiento permite mejorar las políticas de respuesta ante desastres y la estimación del impacto que podrían causar sismos de mayor magnitud en la misma área. Finalmente, y a modo de observación, los sensores sísmicos son artefactos costosos que además implican un alto costo de instalación y mantención, por lo que existen zonas en el mundo que no tienen buena cobertura. Esto puede dificultar la creación de catálogos de sismos suficientemente completos. Por este motivo, se propone la utilización de una fuente de información no convencional, para desarrollar una herramienta de soporte de decisiones durante eventos sísmicos. En particular, la utilización de Twitter, una red social, la cual se caracteriza por sus mensajes de máximo 280 caracteres. Esta red social es utilizada actualmente por millones de usuarios activos que publican mensajes y propagan información en tiempo real, abarcando desde noticias de última hora sobre eventos importantes.Ítem Editorial: Advanced methods in signal processing, image processing and pattern recognition in geosciences(Frontiers Media S.A., 2023-01) Gaci, Said; Nicolis, Orietta; Farfour, MohammedLately, applications of signal processing, image processing, and pattern recognition have been widely introduced in geosciences, such as in the context of natural resources exploration, either petroleum or mineral, and engineering geology (Gaci and Hachay, 2014; Gaci and Hachay, 2017; Gaci et al., 2021). New methods and algorithms have been implemented to discover additional features in different research areas. In petroleum exploration, time-frequency decomposition techniques have been widely applied. The variational mode decomposition was used to detect gas affected by a coal layer in a tight sandstone reservoir located in the Ordos Basin, China. In addition, the multivariate variational mode decomposition was used to accurately estimate the absorption gradient coefficient from data acquired in the Puguang gas field, China.Ítem Etas space–time modeling of chile triggered seismicity using covariates: Some preliminary results(MDPI, 2021-10) Chiodi, Marcello; Nicolis, Orietta; Adelfio, Giada; D’angelo, Nicoletta; Gonzàlez, AlexChilean seismic activity is one of the strongest in the world. As already shown in previous papers, seismic activity can be usefully described by a space–time branching process, such as the ETAS (Epidemic Type Aftershock Sequences) model, which is a semiparametric model with a large time-scale component for the background seismicity and a small time-scale component for the triggered seismicity. The use of covariates can improve the description of triggered seismicity in the ETAS model, so in this paper, we study the Chilean seismicity separately for the North and South area, using some GPS-related data observed together with ordinary catalog data. Our results show evidence that the use of some covariates can improve the fitting of the ETAS model.Ítem Generacion de datos fotométricos artificiales de estrellas variables con Boostrapped GAN(Universidad Andrés Bello, 2022) Vergara Sepúlveda, Álvaro; Peralta, Billy; Nicolis, Orietta; Facultad de IngenieríaThe analysis of astronomical data has made it possible to obtain multiple advances in the understanding of the universe. In astronomy, for example, light curves allow the stars to be characterized, allowing them to be known using the available telescopes. stars are naturally unbalanced by classes which makes automatic recognition difficult as the classification of types of stars. Currently, statistical models have been proposed for the generation of artificial light curves, however these models require assumptions that are not necessarily met in the real data since these models are based on linear relationships that they may not fit non-linear patterns in the actual data. In this work, the generation of artificial data using adversarial generative neural networks is proposed. (GAN) using recurrent networks and considering the generation of time series using bootstrapped sampling of time intervals. The results obtained show that the model is capable of generating visually and quantitatively more realistic photometric data than the obtained by state-of-the-art methods based on parametric statistics. It is concluded that the combination of GAN networks and the bootstrapping method is capable of representing nonlinear and irregular patterns present in real light curves. As future work, we plan to apply attention-based networks to select relevant sections in the generation of artificial light curves using generated and synthetic photometric data.Ítem Métodos de selección de variables óptimas para la predicción de enfermedades cardiovasculares utilizando machine learning(Universidad Andrés Bello, 2020) Rodríguez Segura, Mauricio; Nicolis, Orietta; Facultad de IngenieríaLas enfermedades cardiovasculares (ECV) son la principal causa de muerte en el mundo. La detección temprana de ECV en relación con condiciones del sueño como la apnea y ˜ la actividad física han sido prometedoras y aun es un desafío encontrar nuevas formas de prevenir su aparición. Este trabajo propone metodologías de reducción del número de variables ´ para determinar el riesgo de ECV, mediante métodos de extracción de variables óptimas, ´ con técnicas de pre-procesamiento de datos y evaluando su rendimiento para la clasificación´ predictiva con algoritmos de machine learning (ML) sobre el dataset del Sleep Heart Health Study (SHHS). El pre-procesamiento incluyo el balanceo de datos mediante muestreo SMOTE ´ y la selección de variables óptimas para la predicción de ECV se obtuvo mediante la regresión´ logística con valor p mas bajo y el análisis de componentes principales, utilizando índices médicos y datos de la prueba de polisomnografía. Los algoritmos de ML utilizados para la experimentación fueron: Natıve Bayes (NB), Redes Neuronales Prealimentadas (NN), Maquinas de Soporte Vectorial (SVM) y Bosque Aleatorio (RF). Los resultados obtenidos en ´ el modelo de NN mejoraron la precisión de estudios anteriores (0,81) y presentaron un AUC ´ competitivo (0,76).Ítem Modelo de predicción de riesgo de no rehabilitación cardiovascular con datos limitados usando transfer learning(Universidad Andrés Bello, 2022) Zurita Palma, Christopher; Torres, Romina; Nicolis, Orietta; Facultad de IngenieríaLa rehabilitación cardiovascular, es una etapa que conlleva un conjunto de esfuerzos multidisciplinarios clínicos. A causa de esto, la Fundacion Kaplan en conjunto con la Universidad de Valparaíso y la Universidad Andres Bello, han trabajado y colaborado en ´ el desarrollo de SITECard, una aplicación enfocada a la telerrehabilitación de pacientes en el programa de rehabilitación cardiovascular. Esta plataforma tiene el fin de monitorear los pacientes a distancia, reservar horas, enviar indicaciones remotas a los pacientes, entre otros. En este trabajo nos enfocamos en el componente inteligente de SITECard, el cual consiste en un modelo de aprendizaje automatico de predicción de riesgo de no rehabilitación cardiovascular. Este modelo, fue entrenado con registros retrospectivos de 207 pacientes que participaron en el tratamiento de rehabilitación cardiovascular de la fundación Kaplan, lo que es una cantidad de datos bastante limitada para la obtención de buenos resultados. El desafío que se plantea en este documento, es el de mejorar la precisión de la predicción de este modelo de aprendizaje automatico preexistente (R2 0.716 en el mejor modelo), mediante la incorporación de nuevas características, provenientes de un conjunto de datos biométricos que a su vez fue recolectado de una serie de fichas, exámenes y dispositivos clínicos. Para lograr la mejora del modelo preexistente, se ha utilizado la técnica “JDA” la cual permite realizar una adaptación de características entre conjuntos de datos con distribuciones diferentes. Con su utilización se logró una transferencia de aprendizaje basada en características, entre el conjunto de datos utilizado en el modelo preexistente y el nuevo conjunto de datos biométricos. También se han utilizado técnicas como “RFECV” para la selección de características y “Aprendizaje jerárquico”, para ayudar a lidiar con la limitada data disponible. Mediante la utilización en conjunto de todas estas técnicas se ha logrado mejorar la predicción del modelo de riesgo de no rehabilitación hasta un R2 de un 0.923, en el mejor modelo reportado.Ítem Multi-Objective Location Problem for Bank Branches(Institute of Electrical and Electronics Engineers Inc., 2023) Montero, Elizabeth; Nicolis, Orietta; Reid, Samantha; Torres, MarceloAlthough current technologies allow simple online banking transactions, it is vital to guarantee access to bank branches both for customers and businesses. Banking branch interactions improve trust in bank institutions in everyday situations and emergencies. This research addresses the problem of the location of banking facilities. A set of objective functions related to socio-economic variables, geographical coverage and bankers preferences are considered. A multi-objective linear programming model is proposed to solve different versions of the problem considering a maximum radial distance coverage. The (mono/bi/multi) objective versions of the problem are analyzed by varying the values of the weights of objective functions in the proposed model that is then solved using AMPL Gurobi solver. The case study is carried out in Santiago de Chile with data from a local banking institution. In the results, the bank's initial situation is analyzed to contrast the solutions found later. This analysis is performed considering the mono-objective, bi-objective, and multi-objective nature of the studied problem. In all these cases, we analyzed changes in the structure of solutions as the coverage of the branches increases. The proposed model can quickly obtain feasible solutions according to the needs and available resources, allowing to replicate the analysis in different institutions under similar conditions considering strategically established priorities. © 2013 IEEE.Ítem Outlier Vehicle Trajectory Detection Using Deep Autoencoders in Santiago, Chile(MDPI, 2023-02) Peralta, Billy; Soria, Richard; Nicolis, Orietta; Ruggeri, Fabrizio; Caro, Luis; Bronfman, AndrésIn the last decade, a large amount of data from vehicle location sensors has been generated due to the massification of GPS systems to track them. This is because these sensors usually include multiple variables such as position, speed, angular position of the vehicle, etc., and, furthermore, they are also usually recorded in very short time intervals. On the other hand, routes are often generated so that they do not correspond to reality, due to artifacts such as buildings, bridges, or sensor failures and where, due to the large amount of data, visual analysis of human expert is unable to detect genuinely anomalous routes. The presence of such abnormalities can lead to faulty sensors being detected which may allow sensor replacement to reliably track the vehicle. However, given the reliability of the available sensors, there are very few examples of such anomalies, which can make it difficult to apply supervised learning techniques. In this work we propose the use of unsupervised deep neural network models based on stacked autoencoders to detect anomalous routes in vehicles within Santiago de Chile. The results show that the proposed model is capable of effectively detecting anomalous paths in real data considering validation given by an expert user, reaching a performance of 82.1% on average. As future work, we propose to incorporate the use of Long Short-Term Memory (LSTM) and attention-based networks in order to improve the detection of anomalous trajectories. © 2023 by the authors.Ítem Predicting Cardiovascular Rehabilitation of Patients with Coronary Artery Disease Using Transfer Feature Learning(Multidisciplinary Digital Publishing Institute (MDPI), 2023-02) Torres, Romina; Zurita, Christopher; Mellado, Diego; Nicolis, Orietta; Saavedra, Carolina; Tuesta, Marcelo; Salinas, Matías; Bertini, Ayleen; Pedemonte, Oneglio; Querales, Marvin; Salas, RodrigoCardiovascular diseases represent the leading cause of death worldwide. Thus, cardiovascular rehabilitation programs are crucial to mitigate the deaths caused by this condition each year, mainly in patients with coronary artery disease. COVID-19 was not only a challenge in this area but also an opportunity to open remote or hybrid versions of these programs, potentially reducing the number of patients who leave rehabilitation programs due to geographical/time barriers. This paper presents a method for building a cardiovascular rehabilitation prediction model using retrospective and prospective data with different features using stacked machine learning, transfer feature learning, and the joint distribution adaptation tool to address this problem. We illustrate the method over a Chilean rehabilitation center, where the prediction performance results obtained for 10-fold cross-validation achieved error levels with an NMSE of (Formula presented.) and an (Formula presented.) of (Formula presented.), where the best-achieved performance was an error level with a normalized mean squared error of 0.008 and an (Formula presented.) up to (Formula presented.). The results are encouraging for remote cardiovascular rehabilitation programs because these models could support the prioritization of remote patients needing more help to succeed in the current rehabilitation phase. © 2023 by the authors.Ítem Prediction of intensity and location of seismic events using deep learning(Elsevier B.V., 2021-04) Nicolis, Orietta; Plaza, Francisco; Salas, RodrigoThe object of this work is to predict the seismic rate in Chile by using two Deep Neural Network (DNN) architectures, Long Short Term Memory (LSTM) and Convolutional Neural Networks (CNN). For this, we propose a methodology based on a three-module approach: a pre-processing module, a spatial and temporal estimation module, and a prediction module. The first module considers the Epidemic-Type Aftershock Sequences (ETAS) model for estimating the intensity function, which will be used for estimating the seismic rate on a 1 × 1 degree grid providing a sequence of daily images covering all the seismic area of Chile. The spatial and temporal estimation module uses the LSTM and CNN for predicting the intensity and the location of earthquakes. The last module integrates the information provided by the DNNs for predicting future values of the maximum seismic rate and their location. In particular, the LSTM will be trained using the maximum intensity of the last 30 days as input for predicting the maximum intensity of the next day, and the CNN will be trained on the last 30 images provided by the application of the ETAS model for predicting the probability that the next day the maximum event will be in certain area of Chile. Some performance indexes (such as R2 and accuracy) will be used for validating the proposed models. © 2020 Elsevier B.V.Ítem Prospección de minerales con clasificación supervisada en base a texturas y dimensiones fractales de imágenes satelitales multiespectrales(Universidad Andrés Bello, 2022) González Herrera, Cristian Alex; Nicolis, Orietta; Facultad de IngenieríaEl estudio y uso de imágenes satelitales para prospección de yacimientos mineros ha experimentado un continuo crecimiento, esto debido en parte la disponibilidad de fotografías satelitales que ha permitido el desarrollo de varias técnicas de prospección y clasificación basados principalmente en métodos de combinación, a través de operaciones de pixeles de bandas de imágenes multiespectrales. El propósito principal de esta investigación es predecir cuadrantes en donde pueda ser más probable encontrar yacimientos mineros de la zona de Tiltil, utilizando técnicas de aprendizaje máquina basándose en indicadores de textura y fractales de imágenes multiespectrales. Los algoritmos de Machine Learning han sido progresivamente utilizados en minería para predecir zonas en donde se encuentran yacimientos, para ello se hace uso de: la regresion logística (RL), Naive Bayes (NV), Random Forest (RF) y Redes neuronales artificiales (ANN). Los algoritmoss recibieron como input un conjunto de características basadas en imágenes, como el mapa geológico, y en las diferentes bandas de las imágenes recolectadas por el satélite Sentinel-2A, además de la elevación del terreno, representada también en una imagen. Todas las bandas de la imagen multiespectral son divididas en subimágenes dentro de una cuadrícula, así como de la elevación y el mapa geológico, y a partir de ellas se obtienen indicadores representativos de cada cuadrante, estos indicadores corresponden a las texturas de Haralick (entropía, energía, correlación, etc.) y dimensiones fractales (FD) para superficies, estos últimos se cargan como vectores de características para alimentar los cuatro modelos de aprendizaje supervisado, obteniéndose un yacimiento mineral de cobre u oro en la zona a estudiar, logrando exactitudes entre un 94% ANN y un 95% con RF.Ítem Reconstructing the Quarterly Series of the Chilean Gross Domestic Product Using a State Space Approach(MDPI, 2023-04) Caamaño-Carrillo, Christian; Contreras-Espinoza, Sergio; Nicolis, OriettaIn this work, we use a cointegration state space approach to estimate the quarterly series of the Chilean Gross Domestic Product (GDP) in the period 1965–2009. First, the equation of Engle–Granger is estimated using the data of the yearly GPD and some related variables, such as the price of copper, the exports of goods and services, and the mining production index. The estimated coefficients of this regression are then used to obtain a first estimation of the quarterly GDP series with measurement errors. A state space model is then applied to improve the preliminary estimation of the GDP with the main purpose of eliminating the maximum error of measurement such that the sum of the three-month values coincide swith the yearly GDP. The results are then compared with the traditional disaggregation methods. The resulting quarterly GDP series reflects the major events of the historical Chilean economy.Ítem Redes neuronales ConvLSTM para la predicción de eventos sísmicos en Chile(Universidad Andrés Bello, 2021) González Fuentes, Alex; Nicolis, Orietta; Peralta Márquez, Billy; Facultad de IngenieríaPredecir el riesgo sísmico es importante para poder tomar decisiones con anticipación y evitar efectos catastróficos. En este trabajo se propone un modelo de red neuronal basado en la red Convolucional (CNN) y en la red Long Short Term Memory (LSTM) para predecir el riesgo sísmico en Chile. En particular, se utilizara una red Multi-column ConvLSTM para la predicción del número medio de eventos sísmico mayores a una magnitud de 2, 8 en la escala de Richter, en las regiones de Chile de Coquimbo y la Araucanía entre los años 2010 y 2017. Para este modelo se ocuparon los valores de la función de intensidad estimada a través del modelo ETAS y el desplazamiento acumulado previo a un los eventos sísmicos. Dada las características espaciales y temporales de los datos sísmicos se consideraron matrices de dimensión 20x20 de los últimos 20 días para predecir el número medio de eventos sísmicos del día siguiente en área determinada. De los resultados obtenidos, la red Multi-column ConvLSTM logró tener un coeficiente de determinación de 0, 72 y un MSE más bajo de otras redes.Ítem Redes neuronales espacio-temporal para la predicción de eventos de crímenes(Universidad Andrés Bello, 2020) Esquivel, Nicolás; Nicolis, Orietta; Peralta Márquez, Billy; Facultad de IngenieríaAvances en el aprendizaje de máquina permiten realizar predicciones utilizando múltiples preprocesamientos según las características de la problemática. Los crímenes son una problemática que afecta a gran parte del mundo y siendo un hecho que afecta el índice de calidad de vida en las ciudades. Algunos trabajos han realizado predicciones sobre la categoría de un crimen en base a otras variables, siendo un proceso que no influye demasiado en las operaciones policiales. Por otro lado, existen herramientas de predicción de eventos de crímenes que no logran implementar información a nivel espacio-temporal en profundidad. En este trabajo se proponen tres redes neuronales para la predicción espacio temporales de los eventos de crímenes en diferentes ciudades. Cada red es capaz de reconocer patrones en el tiempo y en el espacio utilizando secuencias de información en mapas geográficos. Se utiliza la red convolucional autoencoder para la ciudad de Chicago, la red convolucional LSTM (Long short-term memory) y la red grafo convolucional LSTM. Estas redes son utilizadas para la predicción de los delitos: robo común para Chicago, hurto/robo en la calle para Baltimore y el robo de vehículo para la ciudad de Santiago. Los resultados de cada red lograron un R2 de 97 % y 86 % para Chicago y Santiago de Chile respectivamente. Para la ciudad de Baltimore se obtuvo una precisión (accuracy) de 86 % en R2. Cada conjunto de red neuronal con ciudad a predecir es relatada en un artículo científico donde cada uno fue aceptado. Las aplicaciones para las ciudades Chicago y Santiago fueron recibidas por conferencias, para el caso de la ciudad de Baltimore, fue publicada en una revista.Ítem Space-Time Prediction of PM2.5 Concentrations in Santiago de Chile Using LSTM Networks(MDPI, 2022-11) Peralta, Billy; Sepúlveda, Tomás; Nicolis, Orietta; Caro, LuisCurrently, air pollution is a highly important issue in society due to its harmful effects on human health and the environment. The prediction of pollutant concentrations in Santiago de Chile is typically based on statistical methods or classical neural networks. Existing methods often assume that historical values are known at a fixed geographic point, such that air pollution can be predicted at a future hour using time series analysis. However, these methods are inapplicable when it is necessary to know the pollutant concentrations at every point of the space. This work proposes a method that addresses the space-time prediction of PM (Formula presented.) concentration in Santiago de Chile at any spatial points through the use of the LSTM recurrent network model. In particular, by considering historical values of air pollutants (PM (Formula presented.), PM (Formula presented.) and nitrogen dioxide) and meteorological variables (temperature, wind speed and direction and relative humidity), measured at fixed monitoring stations, the proposed model can predict PM (Formula presented.) concentrations for the next 24 h in a new location where measurements are not available. This work describes the experiments carried out, with particular emphasis on the pre-processing step, which constitutes an important factor for obtaining relatively good results. The proposed multilayer LSTM model obtained (Formula presented.) values equal to 0.74 and 0.38 in seven stations when considering forecasts of 1 and 24 h, respectively. As future work, we plan to include more input variables in the proposed model and to use attention-based networks. © 2022 by the authors.Ítem Space‑time clustering of seismic events in Chile using ST‑DBSCAN‑EV algorithm(Springer, 2024) Nicolis, Orietta; Delgado, Luis; Peralta, Billy; Díaz, Mailiu; Chiodi, MarcelloChile is one of the most seismic countries in the world especially due to the subduction of the Nazca plate under the South America plate along the Chilean cost. Normally, the spatial distribution of seismic events tends to form spatial and temporal clusters around the main event including both precursor and aftershock events. However, it is very difficult to identify whether an event is a precursor, a main event or an aftershock. In the literature, only some large earthquakes are well described but it does not exist an automatic method to classify them. In this work, we propose a new density based clustering method, called ST-DBSCAN-EV (Space-time DBSCAN with Epsilon Variable), which allows the Epsilon parameter (the radius) to vary depending on the density of the points. The results of the ST-DBSCAN-EV are validated on three important earthquakes with magnitude greater than 8.0 Mw occurred in Chile in the last 20 years, by carrying out a series of experiments considering different combinations of parameters. A comparison with some traditional clustering techniques such as the DBSCAN, ST-DBSCAN, and the K-means has been implemented for assessing the performance of the proposed method. Almost in all cases ST-DBSCAN-EV outperformed traditional ones by providing an F1-Score metric higher than 0.8. Finally, the results of classification are compared with a declustering method. © The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2024.Ítem Temporal Cox Process with Folded Normal Intensity(MDPI, 2022-10) Nicolis, Orietta; Riquelme Quezada, Luis M.; Ibacache Pulgar, GermánIn this work, the case of a Cox Process with Folded Normal Intensity (CP-FNI), in which the intensity is given by (Formula presented.), where (Formula presented.) is a stationary Gaussian process, is studied. Here, two particular cases are dealt with: (i) when the process (Formula presented.) constitutes a family of independent random variables and with a common probability law (Formula presented.), and (ii) the case in which (Formula presented.) is a second order stationary process, with exponential type covariance function. In these cases, we observe that the properties of the Gaussian process (Formula presented.) are naturally transferred to the intensity (Formula presented.) and that very analytical results are achievable from the analytical point of view for the point process (Formula presented.). Finally, some simulations are presented in order to appreciate what type of counting phenomena can be modeled by these cases of CP-FNI. In particular, it is interesting to see how the trajectories show a tendency of the events to be grouped in certain periods of time, also leaving long periods of time without the occurrence of events. © 2022 by the authors.Ítem Visual recognition incorporating features of self-supervised models for the use of unlabelled data(Universidad Andrés Bello, 2021) Díaz Calderón, Gabriel Antonio; Peralta Márquez, Billy; Nicolis, Orietta; Facultad de IngenieríaAutomatic visual object recognition has gained great popularity in the world and is successfully applied to various areas such as robotics, security or commerce using deep learning techniques. Training in machine learning models based on deep learning requires an enormous amount of supervised data, which is expensive to obtain. An alternative is to use semi-supervised models as co-training where the views given by deep networks are differentiated using models that incorporate lateral information from each training object. In this document, we describe and test a co-training model for deep networks, adding as auxiliary inputs to self-supervised network features. The results show that the proposed model managed to converge using a few dozen iterations, exceeding 2 % in precision compared to recent models. This model, despite its simplicity, manages to be competitive with more complex recent works. As future work, we plan to modify deep self-supervised networks to increase diversity in co-training learning.Ítem Wavelet-Based Entropy Measures to Characterize Two-Dimensional Fractional Brownian Fields(MDPI AG, 2020-02) Nicolis, Orietta; Mateu, Jorge; Contreras-Reyes, Javier E.The aim of this work was to extend the results of Perez et al. (Physica A (2006), 365 (2), 282-288) to the two-dimensional (2D) fractional Brownian field. In particular, we defined Shannon entropy using the wavelet spectrum from which the Hurst exponent is estimated by the regression of the logarithm of the square coefficients over the levels of resolutions. Using the same methodology. we also defined two other entropies in 2D: Tsallis and the Renyi entropies. A simulation study was performed for showing the ability of the method to characterize 2D (in this case, α = 2) self-similar processes. © 2020 by the authors.