Modelo de covariância bayesiana para seleção de protfólios de investimentos

The portfolio theory is a field of study devoted to investigate the decision-making by investors of resources. The purpose of this process is to reduce risk through diversification and thus guarantee a return. Nevertheless, the classical Mean-Variance has been criticized regarding its parameters and it is observed that the use of variance and covariance has sensitivity to the market and parameter estimation. In order to reduce the estimation errors, the Bayesian models have more flexibility in modeling, capable of insert quantitative and qualitative parameters about the behavior of the market as a way of reducing errors. Observing this, the present study aimed to formulate a new matrix model using Bayesian inference as a way to replace the covariance in the MV model, called MCB - Covariance Bayesian model. To evaluate the model, some hypotheses were analyzed using the method ex post facto and sensitivity analysis. The benchmarks used as reference were: (1) the classical Mean Variance, (2) the Bovespa index's market, and (3) in addition 94 investment funds. The returns earned during the period May 2002 to December 2009 demonstrated the superiority of MCB in relation to the classical model MV and the Bovespa Index, but taking a little more diversifiable risk that the MV. The robust analysis of the model, considering the time horizon, found returns near the Bovespa index, taking less risk than the market. Finally, in relation to the index of Mao, the model showed satisfactory, return and risk, especially in longer maturities. Some considerations were made, as well as suggestions for further work

Técnicas de conformação de feixe em arranjo de antenas utilizando aprendizagem por reforço

Beamforming is a technique widely used in various fields. With the aid of an antenna array, the beamforming aims to minimize the contribution of unknown interferents directions, while capturing the desired signal in a given direction. In this thesis are proposed beamforming techniques using Reinforcement Learning (RL) through the Q-Learning algorithm in antennas array. One proposal is to use RL to find the optimal policy selection between the beamforming (BF) and power control (PC) in order to better leverage the individual characteristics of each of them for a certain amount of Signal to Interference plus noise Ration (SINR). Another proposal is to use RL to determine the optimal policy between blind beamforming algorithm of CMA (Constant Modulus Algorithm) and DD (Decision Direct) in multipath environments. Results from simulations showed that the RL technique could be effective in achieving na optimal of switching between different techniques.