Advanced search


Knowledge area




Filter by:

Publication type

Authors

Issue Years

Publishers

Origin repository

Access Level

Language

Subject

Select the topics of your interest and receive the hottest publications in your email

2 results, page 1 of 1

Discovery of x-ray emission from young suns in the small magellanic cloud

SERGIY SILICH (2013)

We report the discovery of extended X-ray emission within the young star cluster NGC602a in the Wing of the Small Magellanic Cloud (SMC) based on observations obtained with the Chandra X-ray Observatory. X-ray emission is detected from the cluster core area with the highest stellar density and from a dusty ridge surrounding the HII region. We use a census of massive stars in the cluster to demonstrate that a cluster wind or wind-blown bubble is unlikely to provide a significant contribution to the X-ray emission detected from the central area of the cluster. We therefore suggest that X-ray emission at the cluster core originates from an ensemble of low- and solar-mass pre-main-sequence (PMS) stars, each of which would be too weak in X-rays to be detected individually. We attribute the X-ray emission from the dusty ridge to the embedded tight cluster of the new-born stars known in this area from infrared studies. Assuming that the levels of X-ray activity in young stars in the low-metallicity environment of NGC 602a are comparable to their Galactic counterparts, then the detected spatial distribution, spectral properties, and level of X-ray emission are largely consistent with those expected from low- and solar-mass PMS stars and young stellar objects (YSOs). This is the first discovery of X-ray emission attributable to PMS stars and YSOs in the SMC, which suggests that the accretion and dynamo processes in young, low-mass objects in the SMC resemble those in the Galaxy.

Article

Magellanic Clouds ISM: bubbles HII regions Stars: winds, outflows Stars: pre-main sequence X-rays: stars CIENCIAS FÍSICO MATEMÁTICAS Y CIENCIAS DE LA TIERRA ASTRONOMÍA Y ASTROFÍSICA ASTRONOMÍA Y ASTROFÍSICA

Confiabilidad y rendimiento adaptativo de un almacenamiento en la multi nube con redes neuronales

Reliability and Adaptive performance of multi-cloud storage with neural networks

Israel Rescalvo Anastacio (2021)

El uso de los servicios en la nube se ha incrementado en los últimos años por lo que usuarios, industrias y gobiernos depositan grandes conjuntos de datos en la infraestructura de los proveedores de servicios (CSP, por sus siglas en inglés). No obstante, algunos servicios como el almacenamiento como servicio involucran severos riesgos de accesibilidad, integridad y privacidad de datos. Una solución a dichos inconvenientes está dada por el uso de múltiples CSPs, evitando así que un único CSP cuente con acceso total a la información confidencial. En este trabajo de tesis, se propone el análisis, diseño e implementación de una red apuntadora para configurar un sistema redundante de números residuales con aproximación de rango (AR-RRNS, por sus siglas en inglés) que permite distribuir la información en 𝑛 CSPs y recuperar la misma solo con 𝑘 de ellos. La red neuronal propuesta busca minimizar la probabilidad de pérdida de información y redundancia, ambos objetivos están en conflicto y es fundamental determinar una adecuada configuración de (𝑘, 𝑛) donde 2 ≤ 𝑘 ≤ 𝑛. Asimismo, la red permite seleccionar CSPs específicos y asignarles un segmento del total de la información. La red apuntadora utiliza un modelo codificador decodificador y un sistema de atención entrenados mediante aprendizaje por refuerzo. Esta estructura le permite a la red afrontar cambios en los parámetros de los CSPs como la probabilidad de error. Además, la mayoría de los cálculos se realizan de manera offline. A partir de los resultados, se llevó a cabo un análisis comparativo con diversas versiones de algoritmos genéticos y el algoritmo de ramificación y poda, todos ellos forman parte fundamental del estado del arte en la optimización. El análisis muestra que la red apuntadora es más eficiente en tiempo que los demás algoritmos y genera soluciones similares en calidad al algoritmo genético simple pero la diversidad es menor con respecto a los algoritmos basado en población.

The use of cloud services has increased in recent years as users, industries and governments deposit large data sets in the infrastructure of service providers (CSP). However, some services such as storage as a service involve severe data accessibility, integrity and privacy risks. A solution to these drawbacks is given by the use of multiple CSPs, thus preventing a single CSP from having full access to confidential information. In this thesis work, the analysis, design and implementation of a pointing network is proposed to configure a redundant system of residual numbers with range approximation (AR-RRNS) that allows to distribute the information in 𝑛 CSPs and recover the same only with 𝑘 of them. The proposed neural network seeks to minimize the probability of information loss and redundancy, both objectives are in conflict and it is essential to determine an adequate configuration of 2 ≤ 𝑘 ≤ 𝑛. Likewise, the network allows selecting specific CSPs and assigning them a segment of the total information. The pointer network uses an encoder-decoder model and attention system trained by reinforcement learning. This structure allows the network to face changes in the parameters of the CSPs such as the probability of error. Also, most of the calculations are done offline. Based on the results, a comparative analysis was carried out with different versions of genetic algorithms and the branching and bound algorithm, all of which are a fundamental part of the state of the art in optimization. The analysis shows that the pointing network is more efficient in time than the other algorithms and generates solutions similar in quality to the simple genetic algorithm, but the diversity is lower with respect to population-based algorithms.

Master thesis

almacenamiento en la nube, optimización multi objetivo, Sistema Numérico de Residuo, seguridad, esquemas de compartición de secretos Multi-objective optimization, cloud storage, Residue Number System, security, Secret Sharing Schemes INGENIERÍA Y TECNOLOGÍA CIENCIAS TECNOLÓGICAS TECNOLOGÍA DE LOS ORDENADORES FIABILIDAD DE LOS ORDENADORES FIABILIDAD DE LOS ORDENADORES