TCC - Bacharelado em Ciência da Computação (Sede)
URI permanente para esta coleçãohttps://arandu.ufrpe.br/handle/123456789/415
Navegar
13 resultados
Resultados da Pesquisa
Item Formação de grupos de alunos baseada em múltiplos critérios(2021-05-27) Fiorentino Neto, Giuseppe; Miranda, Péricles Barbosa Cunha de; http://lattes.cnpq.br/8649204954287770; http://lattes.cnpq.br/6288849609186849Group formation is one of the main steps of the collaborative learning. This paper proposes an intelligent method to optimize the group formation process considering multiple criteria: inter-homogeneity, intra-heterogeneity and empathy. The method was evaluated regarding the performance, being compared to the exhaustive and random approaches; And regarding the pedagogical aspect, being compared with random and self-selected methods. The results showed the potential of the proposed method from the computational point of view as well as the pedagogical point of view.Item Evaluation of dimensionality reduction and truncation techniques forword embeddings(2021-03-03) Aoun, Paulo Henrique Calado; Nascimento, André Câmara Alves do; Silva, Adenilton José da; http://lattes.cnpq.br/0314035098884256; http://lattes.cnpq.br/0622594061462533; http://lattes.cnpq.br/1048218441267310The use of word embeddings is becoming very common in many Natural Language Processing tasks. Most of the time, these require computacional resources that can not be found in most part of the current mobile devices. In this work, we evaluate a combination of numeric truncation and dimensionality reduction strategies in order to obtain smaller vectorial representations without substancial losses in performance.Item Teacher-Student aplicado a diferentes modelos de recompensa do ambiente Lunar Lander(2021-07-20) Silva, Kenedy Felipe dos Santos da; Sampaio, Pablo Azevedo; http://lattes.cnpq.br/8865836949700771; http://lattes.cnpq.br/7479192156880225The techniques studied related to learning by reinforcement are becoming more and more common in real world challenges, but one challenge is to reduce the learning time. Currently, the learning time and/or amount of interactions performed by the reinforcement learning agent can result in high costs in applications, as the training of models can consume a lot of time, requiring many interactions between the agent and the task environment. This work seeks to improve learning using a new combination of techniques, the Teacher-Student technique with Reward Shaping. The Teacher-Student technique aims to choose among a set of similar tasks that train for a main task, according to the student’s learning. The Reward Shaping technique, altering the reward to try to accelerate learning, provides more frequent feedback on appropriate behaviors, that is, reports rewards more often. We adapted Teacher-Student algorithms for this combination of techniques, and used the Lunar Lander environment as a case study, using four reward models designed in (ALBUQUERQUE, 2021) for this environment. Experiments were performed running different trainings to compare this approach with training only on the original Lunar Lander (no rewards change), and with the results obtained in (ALBUQUERQUE, 2021) by adopting each of the rewards models individually. The combination of Teacher-Student techniques with Reward Shaping contributed to a new experience in the reinforcement learning area, managing to accelerate the agent’s learning, considering the duration of 600 thousand training steps, reaching the target performance in 2 out of 5 proposals, in addition to of being able to learn better than the original Lunar Lander approach with PPO algorithm.Item Um currículo de aprendizagem por reforço para recompensas modeladas no Lunar Lander(2021-07-19) Albuquerque, Renilson da Silva; Sampaio, Pablo Azevedo; http://lattes.cnpq.br/8865836949700771; http://lattes.cnpq.br/3364503614448061Reinforcement learning is a machine learning paradigm where the agent learns to solve problems interacting with an environment, executing actions in a trial and error sequence. For each action performed, the agent receives a reward from the environment indicating how effective it was in solving the whole problem. The agent’s objective is to maximize the total reward received. However, in some reinforcement learning problems, the agent needs to learn complex tasks receiving uninformative rewards, leading to the credit assignment problem that slows the agent’s training process. Reward shaping and curriculum learning are techniques that can speed up agent training time by separating the problem into smaller tasks to be solved sequentially, applying smaller and informative rewards for each action performed. Lunar Lander is a simplified 2D simulator used as a benchmark for reinforcement learning solutions to the optimization problem on landing control of a lunar module. However, its standard rewards system assigns much more punitive rewards for the use of the engines, not being very constructive for the agent, which can lead to the credit assignment problem. Hence, this work proposes a curriculum using two additional shaped reward models and runs experiments that aim to minimize the Lunar Lander learning time. This work found that both the new models and the curriculum were more effective in training the Lunar Lander agent compared to the standard rewards model.Item Um currículo de aprendizado por reforço para o cenário “Run to Score with Keeper” do Google Research Football Environment(2019-12-10) Silva, Jonatan Washington Pereira da; Sampaio, Pablo Azevedo; Macário Filho, Valmir; http://lattes.cnpq.br/4346898674852080; http://lattes.cnpq.br/8865836949700771; http://lattes.cnpq.br/6846637095187550Reinforcement learning is a group of techniques that allow an agent to interact with a particular environment. Agents observe the state of the environment and perform an action, the action is evaluated through a reward obtained. The agent objective is to maximize this reward. Various issues such as three-dimensional locomotion and electronic games have been addressed by reinforcement learning (KURACH et al., 2019). The Trainament of agents for a soccer game usually has sparse rewards, what slows learning (MATIISEN et al., 2019). One technique that can solve this obstacle is the curriculum learning proposed in (BENGIO et al., 2009). This technique use simplest tasks of the main task and the increase difficult level with the time. In This work we present two curriculum, identified as 5-15-30-50 e 3-10-20-67, for the scenario Run To Score With Keeper of Football Academy. We have shown that curriculums on average achieved better results compared to training only in the main scenario, without curriculum. Curriculum 3-10-20-67 achieved a better result even considering the pattern deviation.Item Graph Embeddings para Node Classification em representação baseada em grafos de frases em linguagem natural(2019) Silva, João Marcos Nascimento da; Lima, Rinaldo José de; http://lattes.cnpq.br/7645118086647340; http://lattes.cnpq.br/5276914899067852Due to the large amount of works developed in the biomedical field and the availability ofhuge databases on biomedical entities, including proteins, genes and viruses, it comesthe need to be able to automatically index such human knowledge bases.Such need has led to the development and computational tools to assist the researcherin the recovery of specific information involving certain proteins and their relations. Inthis context, two of the main problems in the biomedical area involving techniques of Text Mining most investigated are the Named Entity Recognition (NER) and RelationExtraction.This work focuses on the first problem that serves as a basis for the second, i.e., first wehave to identify and classify the entities and then, with the identified/classified entities,identify the existing relations between them, if any. The approach adopted in this paperis based on the recent techniques of supervised/non-supervised learning of deep neural networks, or Deep Learning (DL). In particular, the problem of NER is investigated usingrecent techniques of dense feature representation using DL.At first, the sentences from a biomedical corpus are represented as graphs thanks tothe generation of annotations (metadata) generated automatically by natural language processing tools, such as tokenization, syntactic parsing, etc. These graphs are thenimported into a graph-based database so that various queries submitted to this data base can be optimized in order to extract both lexical and syntactic attributes (or features) ofthe entities (or nodes) present in the graphs. The information generated in the previousstep is used as input Deep Learning-based algorithms called Graph Embedding (GE)that map the representation of graph nodes (entity) in a dense vector representation(vector of real numbers) that has several properties of interest for this search. Finally,such dense representation of features) are employed as input for supervised machine learning algorithms.This work presents an experimental study where some of the existent algorithms of GEare compared, along with several types of sentence representation based on graphs,and their impacts on the task of entity classification (NER), or node classification. Theexperimental results are promising, reaching more than 90% accuracy in the best casesItem Avaliação entre algoritmos de filtragem colaborativa baseada em vizinhança e transferência de conhecimento para CD-CARS(2019) Silva, Guilherme Melo da; Silva, Douglas Véras e; http://lattes.cnpq.br/2969243668455081; http://lattes.cnpq.br/7122596102314881Recommendations in scenarios with the lack of preferences expressed by users is an importantlimitation for Recommendation Systems (RS). Due to this problem, cross-domain RS (CDRS)searches have gained relevance, where collaborative filtering (CF) is one of the most exploitedtechniques in this area. The CD-CARS system shows that the use of contextual information,available in user preferences, can optimize CF neighborhood-based algorithms, a techniquewidely used in multidomain CF. Although they provide accurate recommendations, some neigh-borhood-based algorithms such as the one used in the CD-CARS have the limitation of the useof multi-domains only in the occurrence of user overlap between domains, a non-trivial scenarioin real databases. This work presents a comparative analysis of different recommendation algo-rithms involving collaborative filtering techniques. The CD-CARS’ NNUserNgbr-transClosure(CF neighborhood-based) and Tracer (CF transfer learning-based) algorithms, were used as thebasis for the recommendation algorithms. In the experiments, the CF algorithms were integratedinto the context-aware techniques, addressed in the CD-CARS: Contextual Pre-Filtering andPost-Filtering, being applied on two data sets, formed by two auxiliary domains and one target,with and without overlap between domains. The MAE and RMSE performance metrics wereused to evaluate the algorithms. The results of the experiments showed that the Tracer algorithmpresented better results concerning the NNUserNgbr-transClosure algorithm in all experimentscenarios without user overlap, with and without the use of the Contextual Pre-Filtering or Post-Filtering.Item Uma análise do impacto da experiência prévia com pensamento computacional no desempenho de estudantes em programação no ensino superior(2019) Silva, Emanuel Leite Oliveira da; Falcão, Taciana Pontual da Rocha; http://lattes.cnpq.br/5706959249737319; http://lattes.cnpq.br/5886730483799524This paper aims to study the effect of previous contact with Computational Thinking instudents of higher education courses. Computational Thinking is a skill that aims to de-velop logical thinking and algorithmic thinking on an ongoing and lifelong basis, helpingthem to solve personal and professional life problems using the techniques of computer science. According to research, more than 50% of students in computer courses willdrop out of the course and one of the main reasons is the difficulty in learning and as-similating the basic and advanced concepts of programming, becoming unmotivated.Thus, this work investigated the feasibility of using computational thinking to help thosestudents with programming learning difficulties. Therefore, two student profiles wereidentified, who had contact with Computational Thinking before and after attending Pro-gramming, and questionnaires were applied to evaluate the perspectives they had onthedisciplineanditsbenefit,whethertheuseofComputationalThinkingwasproductiveor not. Two teachers from the UFRPE Computer Degree course were also interviewedto examine their perspective on Computational Thinking on student performance, com-paring students who had contact before and after attending Programming. From thestudents’ perspective, the use of Computational Thinking assists them in cognitive de-velopment, improving logical thinking and algorithmic thinking, and programming learn-ing. Teachers believe that Computational Thinking cognitively prepares students forProgramming, reducing the effort to assimilate the basics and seeing this approach asan improvement for students.Item Classificação de imagens de textura geradas por gráficos de recorrências no problema de pessoas sofrendo ataques epiléticos(2019) Queiroz, Danielly de Moura Borba; Macário Filho, Valmir; http://lattes.cnpq.br/4346898674852080; http://lattes.cnpq.br/7461629772562910Epilepsy is a neurological condition characterized by the occurrence of epileptic seizuresthat recur in variations. These seizures are clinical manifestations of an abnormal dis-charge of neurons, which are cells that make up the brain. Some features make earlydiagnosis of epilepsy a major challenge, even for the most experienced clinicians. Asmedical aid, there are tests such as electroencephalogram (EEG) represented by timeseries widely used in the diagnosis of epilepsy. Time series are present in various areasof study, such as medicine, biology, economics, among others. Your graphics exposehidden patterns and alter data such as texture patterns as well as those that can beused by texture extraction methods. In addition, there are several tools for extractingtime series information, one of which is the hit image, which is currently used to verifythe change of an unsigned pattern. This paper presents a study of texture descriptorsand classifiers in images of healthy and epileptic people generated by recurrence im-ages. The texture descriptors using this study were: Local Binary Models (LBP), LocalPhase Quantification (LPQ) and Gabor Filter Bank. To the best of our knowledge, nostudy has yet been performed, applying these descriptors to base recurrence imagesused in this work. The evaluation is performed through the average hit, precision, recalland f-measure rate resulting from the following classifiers: textit Random Forest, andtextit Support Vector Machine (SVM). The experiments showed that the SVM classi-fier using the LPQ descriptor showed promising results, obtaining 92.1% hit, recall andf-measure mean and for accuracy obtained 92.26%.Item Projeto integrado de redes ópticas de longa distância e Metropolitanas usando algoritmos de inteligência computacional: estudo de caso para o estado de Pernambuco(2017) Nascimento, Jorge Candeias do; Araújo, Danilo Ricardo Barbosa de; http://lattes.cnpq.br/2708354422178489; http://lattes.cnpq.br/8065833426856653Nowadays, several network technologies with different prices and adaptations are appearing in the market. A network topology project involves several metrics; the metrics are used to evaluate a project. In the evaluation we use metrics such as robustness metrics (which help in the network’s ability to recover from a failure), blocking probability and energy consumption. The best way to optimize infrastructure in a network design would be to use the latest technologies, only the most efficient ones, even if such technologies are more expensive. However, of the metrics to be considered in this type of project, one of them is the cost (capital employed). Therefore, it is not always feasible to use the most expensive ones on the market. Many technical issues can help control the metrics of these projects, among which is the network topology (link interconnection). Multiobjective evolutionary algorithms (algorithms inspired by the evolution of the species) have been studied in the state of the art for the conception of network topologies. At the same time, clustering algorithms (algorithms specialized in separating samples into groups) have been used in other types of network studies. This study aimed to make use of computational intelligence algorithms in the construction of a network topology project, using the state of Pernambuco as a case study. In a first stage of the study, a clustering algorithm was used in the division of the state into groups. The intention of this part of the work was to measure the coverage of the network in relation to the entire size of the state, and thus ensure the completeness of the network. In addition, the clustering stage also aimed to propose a cost control model through the merging of different technologies for the network (Passive or active) depending on the function of the network segment. In a second step, an evolutionary multiobjective algorithm was used to compose several network topologies that served the clusters created in the previous step. This algorithm has evolved the various network topologies in order to improve four metrics, Blocking Probability, Cost, Energy Consumption and Algebraic Connectivity. The multiobjective algorithm was designed as a memetic algorithm, and, after a set of executions, the algorithm performances were compared with and without the alteration. The results of the tests, in the first stage, showed that the clustering techniques are quite efficient and adaptable to the proposed goal both in terms of network completeness and cost control. Already in the second stage, or multiobjective search stage, it was verified, through the use of a quality indicator (hypervolume), that there was an improvement of the algorithm in relation to convergence and diversity to the Pareto curve, with the use in its new form as memetic algorithm.