Data: 09 de setembro de 2016
Local: Sala Multiuso CIC
Horário: 14h
Palestrante: VitorFilincowsky Ribeiro (doutorado)
Título: 4D Air Navigation for Efficient Flight Path with Evolutionary Approaches for Brazilian Scenarios
Resumo: The main goal in performance based operations is the management of flight trajectories in order to optimize the operating capability of the aircraft, reducing fuel burn and emissions during each flight phase. At the same time the maximization of airspace efficiency/capacity needs to be addressed considering local airspace requirements and constraints. The arrival coordination task can be viewed as a scenario in which several aircraft approach at a common merging point within optimum time windows, ensured by ATC through in order to eliminate en-route conflicts with an adequate sequencing of the arrivals at specific waypoints along the route. It is proposed a 4D-compliant framework that integrates the individual performance preferences of the landing aircraft and the ATC management procedures. In this stage of the research, the ATC agent should use two methodologies in order to build valid, safe arrival sequencing that decreases and distributes the overall operational costs among the aircraft in a fair and efficient manner: Particle Swarm Optimization (PSO) and Simulated Annealing (SA). Using actual data from Brasilia International Airport (SBBR) to evaluate the effectiveness of the proposed modeling, a case study shows that 77% of the flights are able to accomplish their desired time window flying a CDA procedure using PSO, while 80,77% are able to do so using SA.
Horário: 15h
Palestrante: Hugo Wruck Schneider (doutorado)
Título: A machine learning method to predict long non-conding RNAs in transcriptomes
Resumo: In recent years, a rapidly increasing number of non-coding RNAs (ncRNAs) are being generated by thousands of sequencing projects, which have been creating a huge volume of transcriptome data. Among ncRNAs, long ncRNAs (lncRNAs), often defined as transcripts with length of more then 200 nucleotides and no apparent coding capacity, are still poorly understood. Here, we present a machine learning method to predict lncRNA transcripts, using attributes from their primary structures selected by Principal Component Analysis (PCA) as well as characteristics about ORFs of the transcripts. PCA was used to find the nucleotide patterns with higher variance within the transcriptome of interest, and the supervised machine learning method Support Vector Machine (SVM) was chosen to create the model to predict lncRNAs. This method was applied to three case studies, with human and mouse data, and validated with other organisms.
Organizadora: Profa Célia Ghedini Ralha (Este endereço de email está sendo protegido de spambots. Você precisa do JavaScript ativado para vê-lo.)
Coordenadora dos Seminários de Pós-Graduação Informática 2016-2