Results 241 to 250 of about 241,763 (280)
Some of the next articles are maybe not open access.
2021 American Control Conference (ACC), 2021
It is well known that the extension of Watkins' algorithm to general function approximation settings is challenging: does the “projected Bellman equation” have a solution? If so, is the solution useful in the sense of generating a good policy? And, if the preceding questions are answered in the affirmative, is the algorithm consistent?
Fan Lu +3 more
openaire +1 more source
It is well known that the extension of Watkins' algorithm to general function approximation settings is challenging: does the “projected Bellman equation” have a solution? If so, is the solution useful in the sense of generating a good policy? And, if the preceding questions are answered in the affirmative, is the algorithm consistent?
Fan Lu +3 more
openaire +1 more source
Neural Computing & Applications, 2003
In this paper we introduce a novel neural reinforcement learning method. Unlike existing methods, our approach does not need a model of the system and can be trained directly using the measurements of the system. We achieve this by only using one function approximator and approximate the improved policy from this.
Stephan ten Hagen, Ben Kr�se
openaire +3 more sources
In this paper we introduce a novel neural reinforcement learning method. Unlike existing methods, our approach does not need a model of the system and can be trained directly using the measurements of the system. We achieve this by only using one function approximator and approximate the improved policy from this.
Stephan ten Hagen, Ben Kr�se
openaire +3 more sources
2021
In this paper we focus on reinforcement learning algorithms that are sensitive to risk. The notion of risk we work with is the well-known conditional value-at-risk (CVaR). We describe a faster method for computing value iteration updates for CVaR markov decision processes (MDP). This improvement then opens doors for a sampling version of the algorithm,
Silvestr Stanko, Karel Macek
openaire +1 more source
In this paper we focus on reinforcement learning algorithms that are sensitive to risk. The notion of risk we work with is the well-known conditional value-at-risk (CVaR). We describe a faster method for computing value iteration updates for CVaR markov decision processes (MDP). This improvement then opens doors for a sampling version of the algorithm,
Silvestr Stanko, Karel Macek
openaire +1 more source
IEEE/WIC International Conference on Intelligent Agent Technology, 2003. IAT 2003., 2004
Reinforcement learning is the problem faced by a controller that must learn behavior through trial and error interactions with a dynamic environment. The controller's goal is to maximize reward over time, by producing an effective mapping of states of actions called policy.
null Fei Qian, null Hironori Hirata
openaire +1 more source
Reinforcement learning is the problem faced by a controller that must learn behavior through trial and error interactions with a dynamic environment. The controller's goal is to maximize reward over time, by producing an effective mapping of states of actions called policy.
null Fei Qian, null Hironori Hirata
openaire +1 more source
2020 3rd International Conference on Control and Robots (ICCR), 2020
Mutual learning is an emerging technique for improving performance in machine learning models by allowing functions to learn from each other. In this paper, we present an updated version of Q-learning, improved with the application of mutual learning techniques. We apply this algorithm to a traditional reinforcement learning control problem and compare
Cameron Reid, Snehasis Mukhopadhyay
openaire +1 more source
Mutual learning is an emerging technique for improving performance in machine learning models by allowing functions to learn from each other. In this paper, we present an updated version of Q-learning, improved with the application of mutual learning techniques. We apply this algorithm to a traditional reinforcement learning control problem and compare
Cameron Reid, Snehasis Mukhopadhyay
openaire +1 more source
Fuzzy Q-learning and dynamical fuzzy Q-learning
Proceedings of 1994 IEEE 3rd International Fuzzy Systems Conference, 2002This paper proposes two reinforcement-based learning algorithms: 1) fuzzy Q-learning in an adaptation of Watkins' Q-learning for fuzzy inference systems; and 2) dynamical fuzzy Q-learning which eliminates some drawbacks of both Q-learning and fuzzy Q-learning. These algorithms are used to improve the rule base of a fuzzy controller. >
openaire +1 more source
Proceedings of 6th International Fuzzy Systems Conference, 2002
This paper proposes an adaptation of Watkins' Q-learning (1989, 1992) for fuzzy inference systems where both the actions and the Q-functions are inferred from fuzzy rules. This approach is compared with genetic algorithm on the cart-centering problem, showing its effectiveness.
P.Y. Glorennec, L. Jouffe
openaire +1 more source
This paper proposes an adaptation of Watkins' Q-learning (1989, 1992) for fuzzy inference systems where both the actions and the Q-functions are inferred from fuzzy rules. This approach is compared with genetic algorithm on the cart-centering problem, showing its effectiveness.
P.Y. Glorennec, L. Jouffe
openaire +1 more source
2021
In this chapter, we will do a deep dive into Q-learning combined with function approximation using neural networks. Q-learning in the context of deep learning using neural networks is also known as Deep Q Networks (DQN). We will first summarize what we have talked about so far with respect to Q-learning. We will then look at code implementations of DQN
openaire +1 more source
In this chapter, we will do a deep dive into Q-learning combined with function approximation using neural networks. Q-learning in the context of deep learning using neural networks is also known as Deep Q Networks (DQN). We will first summarize what we have talked about so far with respect to Q-learning. We will then look at code implementations of DQN
openaire +1 more source
2020
Machine learning (ML) is aimed at autonomous extraction of knowledge from raw real-world data or exemplar instances. Machine learning (Barreno et al. in Proceedings of the 2006 ACM symposium on information, computer and communications security, pp 16–25, 2006 [1]) matches the learned pattern with the objects and predicts the outcome.
Nandita Sengupta, Jaya Sil
openaire +1 more source
Machine learning (ML) is aimed at autonomous extraction of knowledge from raw real-world data or exemplar instances. Machine learning (Barreno et al. in Proceedings of the 2006 ACM symposium on information, computer and communications security, pp 16–25, 2006 [1]) matches the learned pattern with the objects and predicts the outcome.
Nandita Sengupta, Jaya Sil
openaire +1 more source
2020
The purpose of the article is to analyze existing approaches of different states and actions spaces representations for Q-learning algorithm for protein structure folding problem, reveal their advantages and disadvantages and propose the new geometric ???state-space??? representation.
openaire +1 more source
The purpose of the article is to analyze existing approaches of different states and actions spaces representations for Q-learning algorithm for protein structure folding problem, reveal their advantages and disadvantages and propose the new geometric ???state-space??? representation.
openaire +1 more source