Algoritmo Q-learning como estratégia de exploração e/ou explotação para metaheurísticas GRASP e algoritmo genético

Techniques of optimization known as metaheuristics have achieved success in the resolution of many problems classified as NP-Hard. These methods use non deterministic approaches that reach very good solutions which, however, don t guarantee the determination of the global optimum. Beyond the inherent difficulties related to the complexity that characterizes the optimization problems, the metaheuristics still face the dilemma of xploration/exploitation, which consists of choosing between a greedy search and a wider exploration of the solution space. A way to guide such algorithms during the searching of better solutions is supplying them with more knowledge of the problem through the use of a intelligent agent, able to recognize promising regions and also identify when they should diversify the direction of the search. This way, this work proposes the use of Reinforcement Learning technique - Q-learning Algorithm - as exploration/exploitation strategy for the metaheuristics GRASP (Greedy Randomized Adaptive Search Procedure) and Genetic Algorithm. The GRASP metaheuristic uses Q-learning instead of the traditional greedy-random algorithm in the construction phase. This replacement has the purpose of improving the quality of the initial solutions that are used in the local search phase of the GRASP, and also provides for the metaheuristic an adaptive memory mechanism that allows the reuse of good previous decisions and also avoids the repetition of bad decisions. In the Genetic Algorithm, the Q-learning algorithm was used to generate an initial population of high fitness, and after a determined number of generations, where the rate of diversity of the population is less than a certain limit L, it also was applied to supply one of the parents to be used in the genetic crossover operator. Another significant change in the hybrid genetic algorithm is the proposal of a mutually interactive cooperation process between the genetic operators and the Q-learning algorithm. In this interactive/cooperative process, the Q-learning algorithm receives an additional update in the matrix of Q-values based on the current best solution of the Genetic Algorithm. The computational experiments presented in this thesis compares the results obtained with the implementation of traditional versions of GRASP metaheuristic and Genetic Algorithm, with those obtained using the proposed hybrid methods. Both algorithms had been applied successfully to the symmetrical Traveling Salesman Problem, which was modeled as a Markov decision process

Aprendizado por Reforço com Valores deInfluência em Sistemas Multi-Agente

We propose a new paradigm for collective learning in multi-agent systems (MAS) as a solution to the problem in which several agents acting over the same environment must learn how to perform tasks, simultaneously, based on feedbacks given by each one of the other agents. We introduce the proposed paradigm in the form of a reinforcement learning algorithm, nominating it as reinforcement learning with influence values. While learning by rewards, each agent evaluates the relation between the current state and/or action executed at this state (actual believe) together with the reward obtained after all agents that are interacting perform their actions. The reward is a result of the interference of others. The agent considers the opinions of all its colleagues in order to attempt to change the values of its states and/or actions. The idea is that the system, as a whole, must reach an equilibrium, where all agents get satisfied with the obtained results. This means that the values of the state/actions pairs match the reward obtained by each agent. This dynamical way of setting the values for states and/or actions makes this new reinforcement learning paradigm the first to include, naturally, the fact that the presence of other agents in the environment turns it a dynamical model. As a direct result, we implicitly include the internal state, the actions and the rewards obtained by all the other agents in the internal state of each agent. This makes our proposal the first complete solution to the conceptual problem that rises when applying reinforcement learning in multi-agent systems, which is caused by the difference existent between the environment and agent models. With basis on the proposed model, we create the IVQ-learning algorithm that is exhaustive tested in repetitive games with two, three and four agents and in stochastic games that need cooperation and in games that need collaboration. This algorithm shows to be a good option for obtaining solutions that guarantee convergence to the Nash optimum equilibrium in cooperative problems. Experiments performed clear shows that the proposed paradigm is theoretical and experimentally superior to the traditional approaches. Yet, with the creation of this new paradigm the set of reinforcement learning applications in MAS grows up. That is, besides the possibility of applying the algorithm in traditional learning problems in MAS, as for example coordination of tasks in multi-robot systems, it is possible to apply reinforcement learning in problems that are essentially collaborative

Uma contribuição ao estudo das categorias internas e de sua proliferação em redes ARTMAP

ART networks present some advantages: online learning; convergence in a few epochs of training; incremental learning, etc. Even though, some problems exist, such as: categories proliferation, sensitivity to the presentation order of training patterns, the choice of a good vigilance parameter, etc. Among the problems, the most important is the category proliferation that is probably the most critical. This problem makes the network create too many categories, consuming resources to store unnecessarily a large number of categories, impacting negatively or even making the processing time unfeasible, without contributing to the quality of the representation problem, i. e., in many cases, the excessive amount of categories generated by ART networks makes the quality of generation inferior to the one it could reach. Another factor that leads to the category proliferation of ART networks is the difficulty of approximating regions that have non-rectangular geometry, causing a generalization inferior to the one obtained by other methods of classification. From the observation of these problems, three methodologies were proposed, being two of them focused on using a most flexible geometry than the one used by traditional ART networks, which minimize the problem of categories proliferation. The third methodology minimizes the problem of the presentation order of training patterns. To validate these new approaches, many tests were performed, where these results demonstrate that these new methodologies can improve the quality of generalization for ART networks

Uma implementação paralela híbrida para o problema do caixeiro viajante usando algoritmos genéticos, GRASP e aprendizagem por reforço

The metaheuristics techiniques are known to solve optimization problems classified as NP-complete and are successful in obtaining good quality solutions. They use non-deterministic approaches to generate solutions that are close to the optimal, without the guarantee of finding the global optimum. Motivated by the difficulties in the resolution of these problems, this work proposes the development of parallel hybrid methods using the reinforcement learning, the metaheuristics GRASP and Genetic Algorithms. With the use of these techniques, we aim to contribute to improved efficiency in obtaining efficient solutions. In this case, instead of using the Q-learning algorithm by reinforcement learning, just as a technique for generating the initial solutions of metaheuristics, we use it in a cooperative and competitive approach with the Genetic Algorithm and GRASP, in an parallel implementation. In this context, was possible to verify that the implementations in this study showed satisfactory results, in both strategies, that is, in cooperation and competition between them and the cooperation and competition between groups. In some instances were found the global optimum, in others theses implementations reach close to it. In this sense was an analyze of the performance for this proposed approach was done and it shows a good performance on the requeriments that prove the efficiency and speedup (gain in speed with the parallel processing) of the implementations performed

Roteamento em Redes de Sensores Sem Fios Com Base Em Aprendizagem Por Reforço

The use of wireless sensor and actuator networks in industry has been increasing past few years, bringing multiple benefits compared to wired systems, like network flexibility and manageability. Such networks consists of a possibly large number of small and autonomous sensor and actuator devices with wireless communication capabilities. The data collected by sensors are sent directly or through intermediary nodes along the network to a base station called sink node. The data routing in this environment is an essential matter since it is strictly bounded to the energy efficiency, thus the network lifetime. This work investigates the application of a routing technique based on Reinforcement Learning s Q-Learning algorithm to a wireless sensor network by using an NS-2 simulated environment. Several metrics like energy consumption, data packet delivery rates and delays are used to validate de proposal comparing it with another solutions existing in the literature

Otimização em comitês de classificadores: uma abordagem baseada em filtro para seleção de subconjuntos de atributos

Traditional applications of feature selection in areas such as data mining, machine learning and pattern recognition aim to improve the accuracy and to reduce the computational cost of the model. It is done through the removal of redundant, irrelevant or noisy data, finding a representative subset of data that reduces its dimensionality without loss of performance. With the development of research in ensemble of classifiers and the verification that this type of model has better performance than the individual models, if the base classifiers are diverse, comes a new field of application to the research of feature selection. In this new field, it is desired to find diverse subsets of features for the construction of base classifiers for the ensemble systems. This work proposes an approach that maximizes the diversity of the ensembles by selecting subsets of features using a model independent of the learning algorithm and with low computational cost. This is done using bio-inspired metaheuristics with evaluation filter-based criteria

Abordagem de construção de arquitetura homogênea para comitês via meta-aprendizagem

In the world we are constantly performing everyday actions. Two of these actions are frequent and of great importance: classify (sort by classes) and take decision. When we encounter problems with a relatively high degree of complexity, we tend to seek other opinions, usually from people who have some knowledge or even to the extent possible, are experts in the problem domain in question in order to help us in the decision-making process. Both the classification process as the process of decision making, we are guided by consideration of the characteristics involved in the specific problem. The characterization of a set of objects is part of the decision making process in general. In Machine Learning this classification happens through a learning algorithm and the characterization is applied to databases. The classification algorithms can be employed individually or by machine committees. The choice of the best methods to be used in the construction of a committee is a very arduous task. In this work, it will be investigated meta-learning techniques in selecting the best configuration parameters of homogeneous committees for applications in various classification problems. These parameters are: the base classifier, the architecture and the size of this architecture. We investigated nine types of inductors candidates for based classifier, two methods of generation of architecture and nine medium-sized groups for architecture. Dimensionality reduction techniques have been applied to metabases looking for improvement. Five classifiers methods are investigated as meta-learners in the process of choosing the best parameters of a homogeneous committee.

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.