deep reinforcement learning

Deep reinforcement learning is widely used in machine learning problems and the use of methods to improve its performance is important. Balance between exploration and exploitation is one of the important issues in reinforcement learning and for this purpose, action selection methods that involve exploration such as ɛ-greedy and Soft-max are used. In these methods, by generating random numbers and evaluating the action-value, an action is selected that can maintain this balance. Over time, with appropriate exploration, it can be expected that the environment becomes better understood and more valuable actions are identified. Chaos, with features such as high sensitivity to initial conditions, non-periodicity, unpredictability, exploration of all possible search space states, and pseudo-random behavior, has many applications. In this article, numbers generated by chaotic systems are used for the ɛ-greedy action selection method in deep reinforcement learning to improve the balance between exploration and exploitation; in addition, the impact of using chaos in replay buffer will also be investigated. Experiments conducted in the Lunar Lander environment demonstrate a significant increase in learning speed and higher rewards in this environment

Recommender systems are one of the most important topics in academia and industry. With the increase in the volume of information and data, it has become confusing and sometimes impossible for users to access the required services without using recommender systems. So far, various techniques have been proposed for this purpose such as collaborative filtering, matrix factorization, logistic regression, neural networks, etc. However, most of these methods suffer from two limitations: (1) considering the recommendation as a static procedure and ignoring the dynamic interactive nature between users and the recommender systems; (2) focusing on the immediate feedback of recommended items and neglecting the long-term rewards. In this research, the modeling of interactions between users and items is done using an improved deep reinforcement learning method which can consider both the dynamic adaptation and long term rewards. The results of the experiments show that the proposed algorithm performs better than other methods.

Computer games have played an important role in the development of artificial intelligence in recent years. Throughout the history of artificial intelligence, computer games have been a suitable test environment for evaluating new approaches and algorithms to artificial intelligence. Different methods, including rule-based methods, tree search methods, and machine learning methods (supervised learning and reinforcement learning) have been developed to create intelligent agents in different games. Games have been used as a suitable environment for trial and error, testing different artificial intelligence ideas and algorithms. Among these researches, we can mention the research of Deep Blue in the game chess and AlphaGo in the game Go. AlphaGo is the first computer program to defeat an expert human Go player. Also, Deep Blue is a chess-playing expert system is the first computer program to win a match, against a world champion. In this paper, we focus on the match-3 game. The match-3 game is a popular game in cell phones, which consists of a very large random state space that makes learning difficult. It also has random reward function which makes learning unstable.  Many researches have been done in the past on different games, including match-3. The aim of these researches has generally been to play optimally or to predict the difficulty of stages designed for human players. Predicting the difficulty of stages helps game developers to improve the quality of their games and provide a better experience for users. Based on the approach used, past works can be divided into three main categories including search-based methods, machine learning methods and heuristic methods. In this paper, an intelligent agent based on deep reinforcement learning is presented, whose goal is to maximize the score in the match-3 game. Reinforcement learning is one of the approaches that has received a lot of attention recently. Reinforcement learning is one of the branches of machine learning in which the agent learns the optimal policy for choosing actions in different spaces through its experiences of interacting with the environment. In deep reinforcement learning, reinforcement learning algorithms are used along with deep neural networks. In the proposed method, different mapping mechanisms for action space and state space are used. Also, a novel structure of neural network for the match-3 game environment has been proposed to achieve the ability to learn large state space. The contributions of this article can be summarized as follow. An approach for mapping the action space to a two-dimensional matrix is presented in which it is possible to easily separate valid and invalid actions. An approach has been designed to map the state space to the input of the deep neural network, which reduces the input space by reducing the depth of the convolutional filter and thus improves the learning process. The reward function has made the learning process stable by separating random rewards from deterministic rewards. The comparison of the proposed method with other existing methods, including PPO, DQN, A3C, greedy method and human agents shows the superior performance of the proposed method in the match-3 game

In this research we develop a deep reinforcement learning-based method for autonomous robot navigation. Our approach in this study is based on DDPG and one of its improved versions named SD3. We did some modifications on this algorithm to make it proper for autonomous navigation problems and optimize it for this problems. The modified algorithm can work with high dimensional state spaces because of using convolutional layers. Also we propose two reward terms include linear velocity reward and angular velocity penalty to encourage robot to move faster with smoother movements. For generalizing the algorithm we used an algorithm for randomly changing shape, layout and number of obstacles in the environment. And to speed up the learning process and improving the robot operation, we normalized all input data. Finally, the proposed algorithm is implemented with ROS and Gazebo and the results show improvement versus the main SD3 and DDPG algorithms.

With the advent and development of IoT applications in recent years, the number of smart devices and consequently the volume of data collected by them are rapidly increasing. On the other hand, most of the IoT applications require real-time data analysis and low latency in service delivery. Under these circumstances, sending the huge volume of various data to the cloud data centers for processing and analytical purposes is impractical and the fog computing paradigm seems a better choice. Because of limited computational resources in fog nodes, efficient utilization of them is of great importance. In this paper, the scheduling of IoT application tasks in the fog computing paradigm has been considered. The main goal of this study is to reduce the latency of service delivery, in which we have used the deep reinforcement learning approach to meet it. The proposed method of this paper is a combination of the Q-Learning algorithm, deep learning, experience replay, and target network techniques. According to experiment results, The DQLTS algorithm has improved the ASD metric by 76% in comparison to QLTS and 6.5% compared to the RS algorithm. Moreover, it has been reached to faster convergence time than QLTS.

To accelerate the learning process in high-dimensional learning problems, the combination of TD techniques, such as Q-learning or SARSA, is usually used with the mechanism of Eligibility Traces. In the newly introduced DQN algorithm, it has been attempted to using deep neural networks in Q learning, to enable reinforcement learning algorithms to reach a greater understanding of the visual world and to address issues Spread in the past that was considered unbreakable. DQN, which is called a deep reinforcement learning algorithm, has a low learning speed. In this paper, we try to use the mechanism of Eligibility Traces, which is one of the basic methods in reinforcement learning, in combination with deep neural networks to improve the learning process speed. Also, for comparing the efficiency with the DQN algorithm, a number of Atari 2600 games were tested and the experimental results obtained showed that the proposed method significantly reduced learning time compared to the DQN algorithm and converges faster to the optimal model.