Horário: 14h
Palestrante:George Oliveira Barros (doutorando)
Orientador: Prof Flavio Vidal
Título: Técnicas em Redes Neurais Convolucionais para Análise de Análise de Imagens de Glomérulos: Podocitopatia
Resumo: As podocitopatias são uma série de lesões que acometem algumas células internas dos glomérulos renais, conhecidas como podócitos. Para identificar essas lesões os patologistas utilizam como padrão ouro, a análise de imagens de biópsia do tecido renal através de microscópio. Esse processo é lento e suscetível à subjetividade. Adicionalmente, detectar os podócitos e os tipos de lesão que os acometem não é uma tarefa trivial, pois necessita de um especialista com experiência e dependendo da imagem, outras células podem ser facilmente confundidas. O objetivo deste trabalho é elaborar técnicas computacionais que realizam a análise de imagens de glomérulos renais com o foco nos podócitos, realizando as tarefas de detecção, classificação e segmentação de lesões podocitárias e seus subtipos.
Horário: 14h20
Palestrante:Daniel Saad Nogueira Nunes (doutorando)
Orientador: Prof Mauricio Ayala Rincón
Title: Grammar Compression by Induced Suffix Sorting
Abstract: A grammar compression algorithm, called GCIS, is introduced in this work. GCIS is based on the induced suffix sorting algorithm SAIS, presented by Nong et al. in 2009. The proposed solution builds on the factorization performed by SAIS during suffix sorting. A context-free grammar is used to replace factors by non-terminals. The algorithm is then recursively applied on the shorter sequence of non-terminals. The resulting grammar is encoded by exploiting some redundancies, such as common prefixes between right-hands of rules, sorted according to SAIS. GCIS excels for its low space and time required for compression while obtaining competitive compression ratios. Our experiments on regular and repetitive, moderate and very large texts, show that GCIS stands as a very convenient choice compared to well-known compressors such as Gzip 7-Zip; and RePair the gold standard in grammar compression; and recent compressors such as SOLCA, LZRR and LZD. In exchange, GCIS is slow at decompressing. Yet, grammar compressors are more convenient than Lempel-Ziv compressors in that one can access text substrings directly in compressed form, without ever decompressing the text. We demonstrate that GCIS is an excellent candidate for this scenario because it shows to be competitive among its RePair based alternatives. We also show that the relation with SAIS makes GCIS a good intermediate structure to build the suffix array and the LCP array during decompression of the text.
Horário: 14h40
Palestrante:Willian de Oliveira Barreiros Júnior (doutorando)
Orientador: Prof George Teodoro
Title: Optimizing Pathology Analysis Pipelines Through Automatic Irregular Image Tiling for Large Scale Distributed Environments
Abstract:The analysis of high resolution whole slide tissue images is a computationally expensive task, which adversely impacts effective use of pathology imaging data in research. We propose runtime solutions to enable efficient execution of pathology image analysis applications on modern distributed memory hybrid platforms equipped with both CPUs and GPUs. Hybrid systems offer significant computation capacity, but taking advantage of this capacity is complex. An application developer has to implement multiple versions of data processing codes targeted for different computing devices. The developer also has to tackle the challenges of efficiently distributing computational load among the nodes of a distributed memory machine and among multiple computing devices on a node. This is a particularly difficult in analysis of high resolution images because of irregular computing costs associated with different image regions. In order to address these problems, we have leveraged a high-level image processing language (Halide) and integrated it into our runtime system called Region Templates (RT). The language simplifies the application development while generating code for multiple devices, such as CPU and GPU. The integration with RT allows for efficient multiple node hybrid execution. We also developed a novel cost-aware data partitioning (CADP) strategy that considers the workload irregularity to minimize load imbalance. Our experimental evaluation shows significant performance improvements on hybrid CPU-GPU machines, as compared with using a single processor (CPU or GPU), as well as on multi-GPU systems. We show that CADP resulted in 1.7× better performance than other workload partitioning approaches (e.g., KD-Trees) on a hybrid machine and was up to 2.24× faster in multi-node settings.
Horário: 15h00
Palestrante:Guilherme Mantovani Oliveira (mestrando)
Orientadora: Profa Priscila Solis
Título: Otimização de Roteamento em Redes SDN (provisório)
Resumo: Modificar protocolos convencionais de roteamento para aplicação em redes SDN.
Horário: 15h20
Palestrante:Guilherme Enéas Vaz Silva (doutorando)
Orientador: Prof André Drummond
Title: Deep-Quality-EON Classifier and DRL approach
Abstract: Elastic Optical Networks (EON) provide better spectrum allocation flexibility and scalability than traditional Wavelength Division Multiplexing networks and are suitable to support the increasing demand for Internet traffic. Those are circuit-switched networks that employ routing, modulation and spectrum allocation (RMSA) algorithms to establish optical circuits. To guarantee an efficient operation of these networks it is necessary to be able to monitor its behavior in a dynamic fashion in search of opportunities to improve its control. In this work, we propose Deep-Quality-EON, a deep learning-based model that classifies the efficiency of the resource allocation strategy applied in an EON solely based a snapshot of the spectrum allocation from the network. The deep learning classifier proposed was evaluated in several scenarios and the results obtained are promising, opening a new avenue on the research area of adaptive RMSA algorithms.
Horário: 15h40
Palestrante:Lucélia Vieira Mota (mestranda)
Orientadora: Prof Thiago Faleiros
Titulo: O uso da supervisão fraca na criação de uma base de entidades rotuladas.
Resumo: Utilizar a técnica de supervisão fraca em base de dados do DODF para geração de uma uma nova base de entidades rotuladas e validar a eficácia dessa técnica comparando os resultados com a base de dados de entidades nomeadas criada por especialistas.
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