Reinforcement learning algorithms: A brief survey

Expert Systems with Applications, Год журнала: 2023, Номер 231, С. 120495 - 120495

Опубликована: Май 23, 2023

Язык: Английский

---

A Survey on Offline Reinforcement Learning: Taxonomy, Review, and Open Problems

IEEE Transactions on Neural Networks and Learning Systems, Год журнала: 2023, Номер 35(8), С. 10237 - 10257

Опубликована: Март 22, 2023

With the widespread adoption of deep learning, reinforcement learning (RL) has experienced a dramatic increase in popularity, scaling to previously intractable problems, such as playing complex games from pixel observations, sustaining conversations with humans, and controlling robotic agents. However, there is still wide range domains inaccessible RL due high cost danger interacting environment. Offline paradigm that learns exclusively static datasets collected interactions, making it feasible extract policies large diverse training datasets. Effective offline algorithms have much wider applications than online RL, being particularly appealing for real-world applications, education, healthcare, robotics. In this work, we contribute unifying taxonomy classify methods. Furthermore, provide comprehensive review latest algorithmic breakthroughs field using unified notation well existing benchmarks' properties shortcomings. Additionally, figure summarizes performance each method class methods on different dataset properties, equipping researchers tools decide which type algorithm best suited problem at hand identify classes look most promising. Finally, our perspective open problems propose future research directions rapidly growing field.

Язык: Английский

---

Challenges and strategies for wide-scale artificial intelligence (AI) deployment in healthcare practices: A perspective for healthcare organizations

Artificial Intelligence in Medicine, Год журнала: 2024, Номер 151, С. 102861 - 102861

Опубликована: Март 30, 2024

Язык: Английский

---

A Systematic Study on Reinforcement Learning Based Applications

Energies, Год журнала: 2023, Номер 16(3), С. 1512 - 1512

Опубликована: Фев. 3, 2023

We have analyzed 127 publications for this review paper, which discuss applications of Reinforcement Learning (RL) in marketing, robotics, gaming, automated cars, natural language processing (NLP), internet things security, recommendation systems, finance, and energy management. The optimization use is critical today’s environment. mainly focus on the RL application Traditional rule-based systems a set predefined rules. As result, they may become rigid unable to adjust changing situations or unforeseen events. can overcome these drawbacks. learns by exploring environment randomly based experience, it continues expand its knowledge. Many researchers are working RL-based management (EMS). utilized such as optimizing smart buildings, hybrid automobiles, grids, managing renewable resources. contributes achieving net zero carbon emissions sustainable In context technology, be optimize regulation building heating, ventilation, air conditioning (HVAC) reduce consumption while maintaining comfortable atmosphere. EMS accomplished teaching an agent make judgments sensor data, temperature occupancy, modify HVAC system settings. has proven beneficial lowering usage buildings active research area buildings. used electric vehicles (HEVs) learning optimal control policy maximize battery life fuel efficiency. acquired remarkable position gaming applications. majority security-related operate simulated recommender provide good suggestions accuracy diversity. This article assists novice comprehending foundations reinforcement

Язык: Английский

---

Deep deterministic policy gradient algorithm: A systematic review

Heliyon, Год журнала: 2024, Номер 10(9), С. e30697 - e30697

Опубликована: Май 1, 2024

Язык: Английский

---

Defining intelligence: Bridging the gap between human and artificial perspectives

Intelligence, Год журнала: 2024, Номер 104, С. 101832 - 101832

Опубликована: Апрель 8, 2024

Achieving a widely accepted definition of human intelligence has been challenging, situation mirrored by the diverse definitions artificial in computer science. By critically examining published definitions, highlighting both consistencies and inconsistencies, this paper proposes refined nomenclature that harmonizes conceptualizations across two disciplines. Abstract operational for are proposed emphasize maximal capacity completing novel goals successfully through respective perceptual-cognitive computational processes. Additionally, support considering intelligence, artificial, as consistent with multidimensional model capabilities is provided. The implications current practices training testing also described, they can be expected to lead achievement or expertise rather than intelligence. Paralleling psychometrics, 'AI metrics' suggested needed science discipline acknowledges importance test reliability validity, well standardized measurement procedures system evaluations. Drawing parallels general (AGI) described reflection shared variance performances. We conclude evidence more greatly supports observation over However, interdisciplinary collaborations, based on common understandings nature sound practices, could facilitate scientific innovations help bridge gap between human-like

Язык: Английский

---

Combining Lyapunov Optimization With Actor–Critic Networks for Privacy-Aware IIoT Computation Offloading

IEEE Internet of Things Journal, Год журнала: 2024, Номер 11(10), С. 17437 - 17452

Опубликована: Янв. 22, 2024

Opportunistic computation offloading is an effective way to improve the computing performance of Industrial Internet Things (IIoT) devices. However, as more and tasks are being offloaded mobile-edge (MEC) servers for processing, it can lead IIoT privacy security issues, such personal usage habits. In this paper, we aim design a Lyapunov-based privacy-aware framework that defines amount user designs "reduced privacy" mechanism. We first define cumulative each trigger protection mechanism when exceeds set threshold. The data generated by then transferred local finally, reduced. This model ensures all users remains stable. further combine advantages Lyapunov optimization actor-critic networks address problem how make learn optimal policy maintain minimum energy consumption in long run. Especially, integrates model-based model-free handle with very low computational complexity, minimizes while stabilizing queue. It demonstrated through experimental simulation results proposed scheme queue stability minimize under strict security.

Язык: Английский

---

Exploring reinforcement learning in process control: a comprehensive survey

International Journal of Systems Science, Год журнала: 2025, Номер unknown, С. 1 - 30

Опубликована: Март 2, 2025

Reinforcement Learning (RL) is a machine learning methodology that develops the capability to make sequential decisions in intricate issues using trial-and-error techniques. RL has become increasingly prevalent for decision-making and control tasks diverse fields such as industrial processes, biochemical systems energy management. This review paper presents comprehensive examination of development, models, algorithms practical uses RL, with specific emphasis on its application process control. The study examines fundamental theories, applications classifying them into two categories: classical Markov decision processes (MDP) deep viz., actor critic methods. topic discussion multiple industries, chemical control, systems, wastewater treatment oil gas sector. Nevertheless, also highlights challenges hinder larger acceptance, including requirement substantial computational resources, complexity simulating real-world settings challenge guaranteeing stability resilience dynamic unpredictable environments. demonstrated significant promise, but more research needed fully integrate it environmental order solve current challenges.

Язык: Английский

---

Robotics in construction: A critical review of the reinforcement learning and imitation learning paradigms

Advanced Engineering Informatics, Год журнала: 2022, Номер 54, С. 101787 - 101787

Опубликована: Окт. 1, 2022

The reinforcement and imitation learning paradigms have the potential to revolutionise robotics. Many successful developments been reported in literature; however, these approaches not explored widely robotics for construction. objective of this paper is consolidate, structure, summarise research knowledge at intersection robotics, learning, A two-strand approach literature review was employed. bottom-up analyse detail a selected number relevant publications, top-down which large papers were analysed identify common themes trends. This study found that on construction has increased significantly since 1980s, terms publications. Also, lacks development dedicated systems, limits their effectiveness. Moreover, unlike manufacturing, construction's unstructured dynamic characteristics are major challenge approaches. provides very useful starting point understating by (i) identifying strengths limitations approaches, (ii) contextualising problem; both will aid kick-start subject or boost existing efforts.

Язык: Английский

---

The Benefits and Limitations of ChatGPT in Business Education and Research: A Focus on Management Science, Operations Management and Data Analytics

SSRN Electronic Journal, Год журнала: 2023, Номер unknown

Опубликована: Янв. 1, 2023

ChatGPT is an artificial-intelligence chatbot developed by OpenAI. It can be used in a variety of applications including content creation, personalized recommendations, copy and for language translation. In Business, it data analysis, provide even process orders. Its benefits have been discussed widely popular media with several articles focusing on the changes will bring to workforce way we live work broadly. this article, discuss limitations Business education research particular focus areas management science, operations analytics. We consider its use both professors students. For professors, design courses, create syllabi content, help grading, student understanding. students, explain complex concepts, debug code, sample exam questions. Overall, find that writing debugging code greatest strength educational purposes. However, has often makes mistakes requires deeper or advanced knowledge domain. Finally, discussion also raises problems regarding bias plagiarism.

Язык: Английский

---