Horário: 14h
Palestrante:Pedro Borges Pio (mestrando)
Orientador: Prof Luis Paulo Faina Garcia
Título: Recomendação de algoritmos de pre-processamento utilizando meta-learning
Resumo: Nos últimos anos, implementou-se um grande esforço para automatizar o processo de aplicação das técnicas de aprendizagem de maquina. Porém, ao mesmo tempo que se obteve um grande avanço nessa área e sendo conhecido que técnicas de pré-processamento possuem um grande impacto no resultado final de um modelo de classificação ou regressão, a automação do pré-processamento ainda pode melhorar bastante. Nesse aspecto, meta-learning pode ser uma boa solução para recomendar um conjunto de melhores algoritmos de pré-processamento para determinado conjunto de dados e modelo de machine learning. Extraindo um conjunto de meta-features, procura-se encontrar padrões que indiquem qual a melhor escolha de algoritimos dentro de um conjunto previamente definido, assim, direcionando melhor as escolhas e otimizando o tempo gasto para realizar o pre-processamento.
Horário: 14h20
Palestrante:Ítalo Barbosa Brasileiro(doutorando)
Orientador: Prof André Drummond
Title: Enabling Sparse Core Switching in Elastic Optical Networks With Spatial Division Multiplexing
Abstract: The single-core fibers applied in the transport network links are taking the bandwidth availability to the limit. As the traffic volume continuously increases year by year, new technologies are demanded to supply future bandwidth requirements. The technology of elastic optical networks emerges as a strong candidate, and the application of multi-core fibers in its links multiplies the availability of resources, by allowing the resource exploration in an extra layer: the spatial domain. However, some adjustments need to be performed, in an attempt to reach the full potential of the multi-core fibers. The core continuity problem causes a division in the SDM-EON literature. The lack of technology to perform core switching reduces the potential of spatial-division multiplexed networks. In this presentation, we evaluate the benefits of core switching occurrence. Furthermore, this presentation evaluates the sparse distribution of nodes with core-switching capabilities, as an attempt to reach a performance near the full core-switch network. To the best of our knowledge, this is the first paper to discuss the sparse core-switching allocation in SDM-EON. Simulations were carried with two distinct topologies and the results evidence that enabling core-switching in only 2 or 3 nodes is sufficient to achieve an efficiency close to a full core-switching network.
Horário: 14h40
Palestrante:Tobias Astoni Sena (doutorando)
Orientador: Prof Vander Alves
Title: Evolution-aware Product-Line Reliability Analysis
Abstract:As any software system, software product lines evolve. Nevertheless, most state-of-the-art product-line analysis techniques do not consider this fact and thus perform analysis from scratch in each evolution step. In the case of reliability analysis, this means that, depending on the evolution scenario, computations for unaffected parts of the software are redone obtaining the same partial results. This wastes computational resources, which is especially problematic since these analyses are time-consuming given the challenge of coping with the state explosion problem compounded with the variability inherent to product lines. We propose a method implemented in the ReAnaE tool to perform incremental product-line reliability analysis, in which analysis results and artifacts are reused whenever possible across the evolution history of the product line.
Horário: 15h00
Palestrante:João Victor de Araujo Oliveira (doutorando)
Orientadora: Profa Maria Emília M. T. Walter
Title: Identification of Telomerase RNA in Yeasts
Abstract: Telomerase RNA (TER) is an essential component of the telomerase ribunocleoprotein process and it varies substantially both in sequence composition and size (from 150 nt to more than 2 kb). In yeasts, TERs is large, usually more than 1000 nt and contains elements that have been extensively studied across several disparate species. However, they are very difficult to detect by homology-based methods even in closer yeast species. This happens because TER sequences evolve rapidly at nucleotide level, are subject to large variations in size, and are highly plastic with respect to their secondary structures. The goal of this work is to apply machine learning technics to identify and better understand the TER in fungi, specially on yeasts. Nevertheless, since we have no many sequences available, and the sequences we have are highly divergent, we started this project trying to look for new TERs in an specific subnarrow group of yeasts, the Candida subgroup. When we have more knowledge about Candida TERs, we would expand our searches in subgroups closer to Candida, expanding, then, the knowledge about TER in several other yeasts species.
Horário: 15h20
Palestrante:Dennis Sávio Martins da Silva (mestrando)
Orientadora: Profa Maristela Holanda
Horário: 15h40
Palestrante:Patrícia Medyna Lauritzen de L. Drumond (doutoranda)
Orientador: Prof Teófilo Emídio de Campos
Local: Teams MS - Equipe PPGI-316415 Seminário, Canal Seminários 1-2021
Profa Célia Ghedini Ralha (Este endereço de email está sendo protegido de spambots. Você precisa do JavaScript ativado para vê-lo.)
Coordenadora Seminários de Pós-Graduação em Informática 1-2021