REINFORCEMENT LEARNING FUNDAMENTALS - LEARNING THROUGH REWARDS AND PUNISHMENTS

Dr. Chithra K · Anirudhan Adukkathayar C · Ismail Keshta · Dr. Haewon Byeon

2023年10月 · Xoffencerpublication

電子書

219

頁

評分和評論未經驗證瞭解詳情

關於本電子書

Reinforcement learning is a subfield within the broader domain of machine learning. The crux of the matter is in selecting the optimal course of action to maximize prospective profitability within a given set of conditions. It is utilized by various software and computers to determine the optimal course of action or action route to effectively respond to a given event. In the process of supervised learning, the training data includes the ground truth, and the model is trained using the correct response. In contrast, in the context of reinforcement learning, the absence of a definitive correct answer is seen. Instead, the reinforcement agent exercises its discretion in selecting the appropriate behaviors required to successfully complete the assigned task. This observation highlights a significant distinction between the two modalities of learning. In supervised learning, the training dataset contains the solution key, enabling the model to be trained using the correct answers directly. In the context of unsupervised learning, the model is trained using erroneous or inaccurate responses. Without access to a training dataset, it is implausible for the system to acquire knowledge by any alternative means. The mathematical impossibility of the situation is evident. Reinforcement learning (RL) is a subfield within the domain of artificial intelligence (AI) that focuses on the examination and analysis of decision-making processes. The objective of this study is to ascertain the optimal approach for individuals to navigate a certain context, with the aim of maximizing the potential outcomes resulting from their endeavors. The data employed in reinforcement learning (RL) is obtained through many machine learning algorithms, each of which acquires knowledge through its distinct iteration of the trial-and-error process. Data is not considered a constituent of the input employed in either supervised or unsupervised machine learning methodologies. Both of these machine learning algorithms are not classified as "supervised." Reinforcement learning is a computational approach that involves the utilization of algorithms to acquire knowledge from previous actions' consequences and afterwards choose the most advantageous path of action. Following each stage, the algorithm is provided with input that aids in evaluating the appropriateness, neutrality, or inaccuracy.

關於作者

Dr. Chithra K, is currently working as an Assistant Professor in NMAMIT NITTE, she holds a PhD in Network Security from Anna University, Chennai. Her research interests encompass interdisciplinary areas, machine learning, and deep learning. With an extensive teaching experience of 11 years in the field of Computer Science and Engineering, Dr. Chithra has contributed significantly. Additionally, she has earned recognition as the holder of multiple patents for her innovative work

Anirudhan Adukkathayar C, is currently an Assistant Professor at NMAM Institute of Technology (NMAMIT), Nitte University. He is also a Research Scholar at Reva University. His research interests span interdisciplinary areas, machine learning, and deep learning. With a teaching experience of 5 years in the field of Computer Science and Engineering, Anirudhan has made significant contributions and holds multiple patents for innovative work.

Ismail Keshta, received his B.Sc. and the M.Sc. degrees in computer engineering and his Ph.D. in computer science and engineering from the King Fahd University of Petroleum and Minerals (KFUPM), Dhahran, Saudi Arabia, in 2009, 2011, and 2016, respectively. He was a lecturer in the Computer Engineering Department of KFUPM from 2012 to 2016. Prior to that, in 2011, he was a lecturer in Princess Nourahbint Abdulrahman University and Imam Muhammad ibn Saud Islamic University, Riyadh, Saudi Arabia. He is currently an assistant professor in the computer science and information systems department of AlMaarefa University, Riyadh, Saudi Arabia. His research interests include software process improvement, modeling, and intelligent systems

Dr. Haewon Byeon, received the DrSc degree in Biomedical Science from Ajou University School of Medicine. Haewon Byeon currently works at the Department of Medical Big Data, Inje University. His recent interests focus on health promotion, AImedicine, and biostatistics. He is currently a member of international committee for a Frontiers in Psychiatry, and an editorial board for World Journal of Psychiatry. Also, He were worked on 4 projects (Principal Investigator) from the Ministry of Education, the Korea Research Foundation, and the Ministry of Health and Welfare. Byeon has published more than 343 articles and 19 books.