Examinando por Materia "Redes neurales (Computadores)"
Mostrando 1 - 20 de 24
- Resultados por página
- Opciones de ordenación
Ítem Algoritmo de compresión para la detección del segmento st de una señal electrocardiográfica fetal basado en la transformada wavelet y umbralización multinivelJiménez Pinto, Gonzalo; Rivas Trujillo, EdwinPresentation of lossless compression algorithm based on the selection of the mother wavelet transform and multilevel thresholding for ST segment detection.Ítem Aplicación de las redes neuronales artificiales en redes de difracción de Bragg para la compensación de ganancia no uniforme en amplificadores ópticos.Cagua Herrera, Jhonatan Mcniven; Montoya Alba, David Esteban; Puerto Leguizamón, Gustavo AdolfoThis work is a compilation of the development and testing of artificial neural networks type multilayer perceptron with backward propagation (Feed - forward backpropagation), trained under the secant method (one - step secant), and selected according to the mean square error , all created using the Matlab program, for the optimization of the characteristics (central frequency and length) of several Bragg diffraction gratings. With the purpose of using them in the compensation of non-uniform gain of the fiber amplifier doped with Erbium (EDFA) of an optical link of eight equidistant carriers located between 195 to 196.4 THz, simulated with the Optisystem 14 program.Ítem Aplicación del algoritmo AdaBoost.RT para la predicción del indice COLCAP y el diseño de un controlador no linealReyes Fajardo, Laura Marcela; Gaona Barrera, Andrés EduardoIn this project, a prediction model of the COLCAP Index and a controller with nonlinear characteristics based on the AdaBoost.RT algorithm with self-adaptive φ were generated. For the case of the prediction is carried out from actions that are part of the Colombian stock exchange and from the values of the actions of the current day the prediction of the index COLCAP of the following day is realized, obtaining a model with 48 Weak Learners and an average percentage error of 1.247%. As for the control case, the inverted pendulum system was stabilized by means of a controller designed from the AdaBoost algorithm, obtaining a model with 39 Weak Learners, Overshoot of 2.08 ° and stabilization time of 4.16s. In both cases it was necessary to tune the parameters of the algorithm according to the problem to be solved.Ítem CACTU: Complemento de actualización urbana para la clasificación de imágenes de alta resolución mediante redes neuronales convolucionales.Peralta Rojas, Laura Michelle; Ariza Buitrago, Ardikary; Cuadros Rangel, William Sneyder; ESPEJO VALERO, OSCAR JAVIERThroughout this document, the development of a complement that allows the identification of physical changes in non-urbanized urbanizable lots is exposed, by means of multi-temporal comparison and coverage classification between two high-resolution images on free geographic-type software such as QGIS. , which allows any user to generate a raster file that defines the probability of constructive changes in a particular area during the period between the images by loading a comparison base image and an image from a later time. , specific case of the application, those areas that went from being an undeveloped lot to presenting some type of construction or vice versa. This plugin performs image processing through supervised learning with convolutional neural networks, where a previous training phase is carried out, based on 60% of the total samples duly labeled in the defined classes (Built and not built). ), to subsequently verify it with the 40% of the remaining samples, generating a correspondence validation of the class that it predicts, with respect to the previously labeled class. Subsequently, with the model already trained, the classification of an image with a later date than the one used in the training is carried out, predicting the probability of whether a pixel corresponds to a construction or not. Finally, the plugin performs a subtraction between the two classifications, the input and the predicted one, thus generating a cover change probability raster. Once the input is generated, the user will be able to make the right decisions regarding the work plan, allowing the establishment of optimal areas and routes for gathering information in the field, focusing resources on areas where changes in real estate dynamics are detected. .Ítem Carta de Erlang B implementada como red neuronal artificialMolina Mosquera, Johan Julian; Coronado, Paulo CésarThis document presents the methodology to implement the non-linear function of Erlang B as a multilayer artificial neural network. The number of channels and the traffic intensity of the system are used as input parameters and the blocking probability or degree of service is used as output data. The proposed model allows the development of the view controller model of the native Android application, in which another variable is entered corresponding to the intensity of traffic per user. The Android app showed results with an error of 0.06%, storing the data to visualize them graphically, thus allowing to simulate and analyze the traffic of fixed telephony and mobile telephony systems.Ítem Clasificador de inventario por medio de una red neuronal convolucional con sincronización a una base de datosAvella Castro, Juan Camilo; Mendéz Hernández, Álvaro Javier; Gaona-Barrera, AndresIn this project, the development of a procedure based on the algorithm is carried out Faster RCNN ResNet 101 for classifying an inventory and recording it in a database, in addition to being able to compare the results obtained, adatabase portion for post-validation. The Faster RCNN ResNet 101 algorithm has been implemented in detection problems of objects, obtaining good results with models of low complexity and medium complexity, which is why there has been a growing interest in applying this algorith in detection and classification of objects. Furthermore, no application of the algorithm is presented in Store inventory problems, which is why it is the reason for this project. To carry out the classification of products, it is necessary to carry out the collection of a database, taking into account the division for training, validation and post-validation. This database is obtained by means of two professional cameras by labeling the images with Labelimg and processing the images to increase the information provided by the database. The input parameters of the object detection system are then selected. With base 70% of the database, the training and validation of the model is carried out, for through convolutional networks and networks of relationship proposals, through the model Pretrained Faster RCNN ResNet 101. A variation of an own parameter "learning coefficient" is carried out to find the classifier model. The results show that it is possible to perform a detection of objects through the pretrained model with an approximate performance of 96.2%, but it is necessary to make a database with possible human failures, such as the greater amount of noise in order to obtain an efficient model, since the performance of the model has a direct relationship with the noises in the database. Regarding the counting of the products, a module was generated at the output of the object detection, which shows the number of products per class with an efficiency of the 88.6%, Finally the data was uploaded to FireBase, a dynamic database with various connection tools on platforms.Ítem Deep Learning aplicado a imágenes satelitales como herramienta de detección de viviendas sin servicio de energía en el caserı́o Media Luna-Uribia-GuajiraValdés Ávila, Lalita Sakhi; Baquero Vanegas, Joher Mauricio; Romero Villalobos, Oswaldo AlbertoIn this thesis project, a Deep Learning application is developed, specifically a tool for detection of homes without utility services on the satellite map of the village of Media Luna, located in the municipality of Uribia, north of La Guajira. Using a dataset composed of satellite images of homes in different rural areas of Colombia, obtained through Google Earth, two different prediction models are developed, a comparison of these models is made with the aim of minimizing the prediction error. Different technologies were used to solve the problem, including TensorFlow and Keras for the creation of neural networks, with their respective configurations. Convolutionary Neural Networks are proposed and a pre-trained Keras model called VGG16 with a ReLu activation function. The experiments carried out show that the use of Convolutional Networks and the algorithms presented have an acceptable and more efficient performance than the traditional methods applied for the VSS counting in rural areas, with reasonable processing times and speed in the delivery of the required information.Ítem Desarrollo de un modelo predictivo basado en Machine Learning para observar el comportamiento de los precios de los Tubérculos en la Ciudad de BogotáRiveros Rey, Oscar Arley; Baròn Velandia, JulioThe behavior of prices of tubers prices can be projected from prediction techniques in this document, a price prediction model based on Machine Learning techniques is proposed. The main objective is to establish a price approximation for the city of Bogotá applying as as a statistical method, time series that allow determining price increases or decreases in the commercialization of tubers; as a result of the execution of the model Applied to the tuber (Arracacha) in the first week of December 2018 a deviation of 19,11 was obtainedÍtem Desarrollo e implementación de un método para la clasificación de imágenes alteradas, usando redes neuronales convolucionales y detección de ruido y bordes.Lagos Ruiz, Andrés Felipe; Gaona Barrera, Andrés EduardoToday it is necessary the truth in the information at the moment of public or shares with others. Some ways to change the meaning of an image is with the use of any tool to change some pixels of the image, one way to know if an image has any forgery is with the use of computational intelligence and fake analysis areas. The proposed model consists in a preprocessing of the image to have the same size to implement gaussian noise estimation and sharpness detection. Finally, the sharpness and Gaussian noise is applied to the convolutional neural network to predict if the image has any forgery. The results of the convolutional neural network have a prediction of 89.54. With the detection of sharpness and noise is 90.54 of all of the images.Ítem Diseño y simulación de un control neuronal aplicado a un convertidor flyback para la regulación de tensiónGómez Buitrago, Juan David; López Manchola, Oscar Eduardo; Gaona Barrera, Andrés Eduardo; Díaz Aldana, Nelson LeonardoThis article explains the design and simulation of a controller based on neural networks to regulate the output voltage of a flyback converter. Neural networks are used since they do not require a mathematical model of the converter, with the advantage of a greater operating range than traditional control methods. In the training process, changes were made in the database and in the neural network architecture to get a more appropriate controller that the guaranteed line and load regulation of the converter. The functional neural controller validation was made on Simulink with the circuital model of a flyback converter, putting it to changes of output load and input voltage. The results obtained show the effectiveness of neuronal control with its ability to regulate lines in a range of 20V to 50V, load regulation between 8Ω and 12Ω, and whose architecture is made up of four neurons.Ítem Un estudio comparativo entre ANFIS, ANNs y SONFIS para series temporales volátilesPerdomo Tovar, Jairo Andrés; Galindo Arévalo, Eiber Arley; Figueroa García, Juan CarlosThis paper presents a comparison among ANFIS, ANNs, and a Self Organized Neuro Fuzzy Inference System (SONFIS) for time series prediction. The Turkish stock index (ISE) series is analyzed using the three methods, a statistical analysis of the residuals per method is performed, and the advantages/disadvantages per method are discussed.Ítem Evaluación de redes convolucionales para la segmentación de objetos geográficos: un insumo para la cartografía básica a escala 1:2000 basado en el catálogo del IGACForero Zapata, Sebastian; Herrera Escorcia, José LuisThis research work explores the use of convolutional neural networks (CNNs) for the automatic segmentation of geographic objects in the generation of basic cartography at a scale of 1:2000, focusing on municipalities in Colombia. The similarities between geographic objects (roads, green areas, forests, water bodies, buildings) and the physical characteristics of the region are analyzed, aligning with the IGAC object catalog. The CNN architectures UNet, DeepLabV3, and LinkNet were selected and evaluated, implementing Transfer Learning in UNet. Data from the IGAC was collected and selected, creating a training dataset and performing preprocessing. The performance of the architectures was evaluated using metrics such as Test Loss, IoU, F1, precision, accuracy, and recall. The results indicated that UNet with Transfer Learning achieved the best overall performance, excelling in IoU, F1, precision, accuracy, and recall. It is important to consider practical factors such as training time and adaptability to new data when choosing the most suitable architecture.Ítem Evaluación de una metodología para la generación de modelos digitales del terreno a partir de estereofotogrametría digital aplicando el modelo de correlación por mínimos cuadrados mediante una red neuronal artificialCajamarca Montoya, Yeison Javier; Barragán, WilliamNowadays on the development of a variety of huge projects is necessary in the engineering field, more specifically in the Geomatic field, automation of many processes with high accuracy levels because of that the modeling of all factors zone could be more realistic as possible. This document makes a general approach to three fundamental topics to work on a project which overall objective is to evaluate a methodology for digital terrain model (DTM) production adding the artificial neural network (ANN) theory by the least squares model correlation. The high accuracy DTM development becomes an important data input in the implementation of risk mitigation projects by natural phenomena either land use planning or watersheds behavior; and so on, there are many other knowledge fields that detailed land measure of the area in a large scales provides characteristics that are the basis for specific analyzes on each one.Ítem Exploración del modelo Word2Vec: Bag-of-Words y Skip-gram, en el marco del Procesamiento del Lenguaje Natural.Casas Peñarete , Cristian Camilo; Masmela Caita, Luis Alejandro; Masmela Caita,Luis Alejandro [0000-0003-3882-4980]Our objective is to understand the functioning of the two neural networks that are part of the Word2Vec model, presented by Mikolov et al. To achieve this, we will provide a brief introduction to the concept of neural networks and their training process. Then, we will offer a brief contextualization of natural language processing, and finally, delve into the details of how the Word2Vec models work. This analysis will include notions of why these models are effective, supported by illustrative examples.Ítem Generación y simulación de un modelo predictivo para prevenir inundaciones en viviendas aledañas a zonas de riesgo mediante técnicas de inteligencia artificialMoreno Castillo, Jenny Marcela; Sánchez Céspedes, Juan Manuel; Espitia Cuchango, Helbert EduardoFloods represent one of the disasters that causes most of human and economic losses worldwide. Therefore, in this project the use of several artificial intelligence techniques are proposed with the aim of predicting the water level in Magdalena River, where millions of inhabitants have their houses. For the development of the project, variables such as: historical data of the water level in different seasons, quarter of the year, rainy season and presence or absence of the El Niño-Southern Oscillation (ENSO) phenomenon were used. The results of the MSE error showed a good performance of the different techniques, being the best one Artificial Neural Networks. However, in addition to having a low error level, the PSO technique offered the possibility to interpret the conditions that can trigger a flood.Ítem Implementación de un modelo predictor para la toma de decisiones en redes inalámbricas de radio cognitivaLópez Sarmiento, Danilo Alfonso; Rivas Trujillo, EdwinThe stage of spectral decision making in cognitive radio networks (CRNs) with centralized topologies depends, among other variables, on the reliability of the characterization model of primary users (PUs), the base station request processing method (BS ) And the channel selection algorithm; according to (Masonta, Mzyece, & Ntlatlapa, 2013), (López, Trujillo, & Gualdron, 2015), among other authors, it is necessary to propose and/or apply methodologies that better estimate the presence/ absence of the PUs in the licensed channels, perfect the way in which requests are processed in the BS and improve the sub-stage of selection and allocation of channels in the CRN. In this regard, the doctoral thesis proposes: 1) the use of LSTM, SVM and ANFIS-GRID-FCM to predict the behavior of PUs, 2) poses the possibility of managing requests from secondary users in the BS (using MLPNN) aiming at reducing the time needed for spectrum allocation, 3) generate efficient spectrum selection schemes based on spectrum classification from SVM and ANFIS learning techniques. The methodology used to evaluate/validate the algorithms that make up the decision making system includes as a source of information, the use of a database containing the spectral behavior of PUs in different channels in the licensed GSM band and free WiFi, and generating traffic simulation using QoS criteria for SUs; the programming languages used for building the algorithms are based on the use of C #, Java Script and Matlab. Results show: 1) a higher percentage of success in the LSTM and ANFIS-GRID-FCM characterization, 2) a decrease in the time needed to select and allocate channels, using a proactive strategy to manage SUs requests In relation to those existing in the state of the art, 3) furthermore it proves that SVM and ANFIS are valid for use in channel selection techniques. In conclusion, the proposed decision-making system should be considered as an additional contribution to improve the spectral decision stage in infrastructure-based CRNs, which should be improved/complemented, including such important factors as the characterization of secondary users, the generation of schemas that allow auto configuration of cognitive nodes, and integration of other CR stages, such as spectral mobility in order to determine the feasibility of its implementation on a real scale.Ítem Modelo de predicción de movimiento que emula patrones de acción fijos (PAF), aplicado a brazo con 3 grados de movimientoRey Lancheros, Diego Enrique; Peña Suesca, Rafael AntonioIn this article we propose a model for movement prediction for an arm with 3 degrees of freedom which follows a point in the plane. This model is based on the human movement system which does not occur in real time and consists of fixed action patterns. A spatial representation of the arm is made, also a set of movement rules are defined and finally are fed to a neural network and finish, the algorithm’s operation is explained.Ítem Modelo inteligente de decisión de espectro que mejora el desempeño en redes de radio cognitivaBernal Ariza, Cristian Camilo; Hernández Suárez, Cesar AugustoThis document proposes the design of a dynamic decision-making model in cognitive wireless networks that allows secondary users to opportunely harness the spectrum and use channels without affecting the traffic of primary users. The implemented model includes an initial decision-making algorithm with multiple criteria that classifies the best channels according to the spectrum characterization and a final algorithm for spectral occupancy prediction that allows the secondary user to change channels when the current channel is requested by a primary user. This algorithm was chosen after evaluating three prediction techniques. This work manages to integrate two decision methods that contribute to reducing the amount of channel changes that the secondary user must perform. The analysis of the results from prediction techniques indicate that the Grey Rational Analysis (GRA) algorithm in combination with the Support Vector Machine (SVM) algorithm presents the best performance in terms of choosing an available channel, reducing the primary user’s interference and diminishing the rate of necessary handoffs.Ítem Optimización del control de calidad para el reconocimiento de piezas en células flexibles de manufactura mediante la implementación de redes neuronales.Parra Franco, Carlos Humberto; Torrejano Munévar, Juan Sebastián; Montiel Ariza, Holman; Montaña Quintero, HenryThe high levels of automation in the companies that use manufacturing cells for their fabrication processes guarantee a high efficiency in production with high quality standards. The main objective of this paper is to develop an optimization for the tasks performed by the quality control station through artificial vision belonging to the manufacturing cell FMS-200; the development of the proposal focuses on optimizing the inspection process and control of quality by adapting image processing techniques and adaptive resonance neural networks. For this particular case, a review of the location of the optical system and a detailed time study for each movement of the station actuators was carried out, attempting to determinate a better position for the optical detection system (artificial vision camera) and the possibility elimination of some components of the station; all of this in order to considerably reduce the operating time and increasing the efficiency for the quality control of the system.Ítem Predicción del consumo del ancho de banda de las aplicaciones web en la nube nativa basada en machine learningOsorio Diaz, Ramiro; Ferro Escobar, RobertoIn this research, a comparative study of three neural network algorithms, which allow modeling a multilayer neural network, with a minimum of three layers ; selecting one, whose objective is to learn to predict the measurement of network traffic, which is connected to the cloud to validate the behavior of the network parameter of "bandwidth consumption", to optimize in time the network resources and ensure the improvement of the quality of service of web applications for small and medium enterprises. In recent years artificial neural networks have been used for predictive analysis, which outlines (Piedra et al., 2008), "Thus, some ANN models Artificial Neural Networks are used to determine projections from a data source; this feature can be exploited to make predictions, for example, to determine available bandwidth". Having an overview of the traffic flowing through the network allows to generate a network capacity planning when managing limited resources as in the case of small and medium enterprises, likewise (Piedra et al., 2008) justify the need to predict network traffic, "a traffic prediction system is required for planning and sizing purposes, this will allow to forecast traffic demands according to previous time periods".