• 한국지능시스템학회:학술대회논문집
• /
• 한국퍼지및지능시스템학회 1998년도 추계학술대회 학술발표 논문집
• /
• pp.420-426
• /
• 1998

Evolutionary algorithms are probabilistic optimization algorithms based on the model of natural evolution. Recently the efforts to improve the performance of evolutionary algorithms have been made extensively. In this paper, we introduce the research for improving the convergence rate and search faculty of evolution algorithms by using reinforcement learning. After providing an introduction to evolution algorithms and reinforcement learning, we present adaptive genetic algorithms, reinforcement genetic programming, and reinforcement evolution strategies which are combined with reinforcement learning. Adaptive genetic algorithms generate mutation probabilities of each locus by interacting with the environment according to reinforcement learning. Reinforcement genetic programming executes crossover and mutation operations based on reinforcement and inhibition mechanism of reinforcement learning. Reinforcement evolution strategies use the variances of fitness occurred by mutation to make the reinforcement signals which estimate and control the step length.

• International Journal of Internet, Broadcasting and Communication
• /
• 제15권3호
• /
• pp.166-175
• /
• 2023

Reinforcement learning can be applied to a wide variety of problems. However, the fundamental limitation of reinforcement learning is that it is difficult to derive an answer within a given time because the problems in the real world are too complex. Then, with the development of neural network technology, research on deep reinforcement learning that combines deep learning with reinforcement learning is receiving lots of attention. In this paper, two types of neural networks are combined with reinforcement learning and their characteristics were compared and analyzed with existing value-based reinforcement learning algorithms. Two types of neural networks are FNN and CNN, and existing reinforcement learning algorithms are SARSA and Q-learning.

• Journal of Information Processing Systems
• /
• 제16권5호
• /
• pp.1001-1007
• /
• 2020

In recent years, machine learning algorithms are continuously being used and expanded in various fields, such as facial recognition, signal processing, personal authentication, and stock prediction. In particular, various algorithms, such as deep learning, reinforcement learning, and Q-learning, are continuously being improved. Among these algorithms, the expansion of deep learning is rapidly changing. Nevertheless, machine learning algorithms have not yet been applied in several fields, such as personal authentication technology. This technology is an essential tool in the digital information era, walking recognition technology as promising biometrics, and technology for solving state-space problems. Therefore, algorithm technologies of deep learning, reinforcement learning, and Q-learning, which are typical machine learning algorithms in various fields, such as agricultural technology, personal authentication, wireless network, game, biometric recognition, and image recognition, are being improved and expanded in this paper.

• 한국정보통신학회:학술대회논문집
• /
• 한국정보통신학회 2021년도 추계학술대회
• /
• pp.693-696
• /
• 2021

강화학습에는 다양한 알고리즘이 있으며 분야에 따라 사용되는 알고리즘이 다르다. 게임 분야에서도 강화학습을 사용하여 인공지능을 개발할 때 특정 알고리즘이 사용된다. 알고리즘에 따라 학습 방식이 다르고 그로 인해 만들어지는 인공지능도 달라진다. 그러므로 개발자는 목적에 맞는 인공지능을 구현하기 위해 적절한 알고리즘을 선택해야 한다. 그러기 위해서 개발자는 알고리즘의 학습 방식과 어떤 종류의 인공지능 구현에 적용되는 것이 효율적인지 알고 있어야 한다. 따라서 이 논문에서는 게임 인공지능 구현에 사용되는 알고리즘인 SAC, PPO, POCA 세 가지 알고리즘의 학습 방식과 어떤 종류의 인공지능 구현에 적용되는 것이 효율적인지 비교한다.

• 한국지능시스템학회:학술대회논문집
• /
• 한국퍼지및지능시스템학회 1998년도 춘계학술대회 학술발표 논문집
• /
• pp.81-84
• /
• 1998

Genetic Algorithms (GAs) are stochastic algorithms whose search methods model some natural phenomena. The procedure of GAs may be divided into two sub-procedures : Operation and Selection. Chromosomes can produce new offspring by means of operation, and the fitter chromosomes can produce more offspring than the less fit ones by means of selection. However, operation which is executed randomly and has some limits to its execution can not guarantee to produce fitter chromosomes. Thus, we propose a method which gives a directional information to the genetic operator by reinforcement learning. It can be achived by using neural networks to apply reinforcement learning to the genetic operator. We use the amount of fitness change which can be considered as reinforcement signal to calcualte the error terms for the output units. Then the weights are updated using backpropagtion algorithm. The performance improvement of GAs using reinforcement learning can be measured by applying the pr posed method to GA-hard problem.

• 한국인터넷방송통신학회논문지
• /
• 제20권1호
• /
• pp.171-176
• /
• 2020

강화 학습은 인공지능 에이전트가 비디오 게임을 학습할 때 가장 효과적으로 사용되는 방법이다. 강화 학습을 위해 여지껏 많은 알고리즘들이 제시되어 왔지만 알고리즘마다 적용되는 분야에 따라 다른 성능을 보여주었다. 본 논문은 최근 강화 학습에서 주로 사용되는 알고리즘들의 성능이 2D 레이싱 게임에서 어떻게 달라지는지 비교 평가한다. 이를 위해 평가에서 사용할 성능 메트릭을 정의하고 각 알고리즘에 따른 메트릭의 값을 그래프로 비교하였다. 그 결과 ACER (Actor Critic with Experience Replay)를 사용할 경우 게임의 보상이 다른 알고리즘보다 평균적으로 높은 것을 알 수 있었고, 보상 값이 가장 낮은 알고리즘과의 차이는 157%였다.

• 한국정보처리학회논문지
• /
• 제6권3호
• /
• pp.633-640
• /
• 1999

Reinforcement learning performs learning through interacting with trial-and-error in dynamic environment. Therefore, in dynamic environment, reinforcement learning method like Q-learning and TD(Temporal Difference)-learning are faster in learning than the conventional stochastic learning method. However, because many of the proposed reinforcement learning algorithms are given the reinforcement value only when the learning agent has reached its goal state, most of the reinforcement algorithms converge to the optimal solution too slowly. In this paper, we present COMREL(COMpressed REinforcement Learning) algorithm for finding the shortest path fast in a maze environment, select the candidate states that can guide the shortest path in compressed maze environment, and learn only the candidate states to find the shortest path. After comparing COMREL algorithm with the already existing Q-learning and Priortized Sweeping algorithm, we could see that the learning time shortened very much.

• 제어로봇시스템학회논문지
• /
• 제7권10호
• /
• pp.819-826
• /
• 2001

Genetic algorithms are becoming more popular because of their relative simplicity and robustness. Genetic algorithms are global search techniques for nonlinear optimization. However, traditional genetic algorithms, though robust, are generally not the most successful optimization algorithm on any particular domain because they are poor at hill-climbing, whereas simulated annealing has the ability of probabilistic hill-climbing. Therefore, hybridizing a genetic algorithm with other algorithms can produce better performance than using the genetic algorithm or other algorithms independently. In this paper, we propose an efficient hybrid optimization algorithm named the adaptive random signal-based learning. Random signal-based learning is similar to the reinforcement learning of neural networks. This paper describes the application of genetic algorithms and simulated annealing to a random signal-based learning in order to generate the parameters and reinforcement signal of the random signal-based learning, respectively. The validity of the proposed algorithm is confirmed by applying it to two different examples.

• Geomechanics and Engineering
• /
• 제11권6호
• /
• pp.787-803
• /
• 2016

DDARF (Discontinuous Deformation Analysis for Rock Failure) is a numerical algorithm for simulating jointed rock masses' discontinuous deformation. While its reinforcement simulation is only limited to end-anchorage bolt, which is assumed to be a linear spring simply. Here, several new reinforcement modes in DDARF are proposed, including lining reinforcement, full-length anchorage bolt and equivalent reinforcement. In the numerical simulation, lining part is assigned higher mechanical strength than surrounding rock masses, it may include multiple virtual joints or not, depending on projects. There must be no embedding or stretching between lining blocks and surrounding blocks. To realize simulation of the full-length anchorage bolt, at every discontinuity passed through the bolt, a set of normal and tangential spring needs to be added along the bolt's axial and tangential direction. Thus, bolt's axial force, shearing force and full-length anchorage effect are all realized synchronously. And, failure criterions of anchorage effect are established for different failure modes. In the meantime, from the perspective of improving surrounding rock masses' overall strength, a new equivalent and tentative simulation method is proposed, it can save calculation storage and improve efficiency. Along the text, simulation algorithms and applications of these new reinforcement modes in DDARF are given.

• Nuclear Engineering and Technology
• /
• 제54권10호
• /
• pp.3864-3877
• /
• 2022

We propose a new approach called PESA (Prioritized replay Evolutionary and Swarm Algorithms) combining prioritized replay of reinforcement learning with hybrid evolutionary algorithms. PESA hybridizes different evolutionary and swarm algorithms such as particle swarm optimization, evolution strategies, simulated annealing, and differential evolution, with a modular approach to account for other algorithms. PESA hybridizes three algorithms by storing their solutions in a shared replay memory, then applying prioritized replay to redistribute data between the integral algorithms in frequent form based on their fitness and priority values, which significantly enhances sample diversity and algorithm exploration. Additionally, greedy replay is used implicitly to improve PESA exploitation close to the end of evolution. PESA features in balancing exploration and exploitation during search and the parallel computing result in an agnostic excellent performance over a wide range of experiments and problems presented in this work. PESA also shows very good scalability with number of processors in solving an expensive problem of optimizing nuclear fuel in nuclear power plants. PESA's competitive performance and modularity over all experiments allow it to join the family of evolutionary algorithms as a new hybrid algorithm; unleashing the power of parallel computing for expensive optimization.