Global ETD Search
Search theses and dissertations gathered from participating repositories worldwide. Every result links back to the library that holds it. No account is needed.
Results
Showing 1 to 20 of 57 for “"Offline reinforcement learning"”.
-
Federated Offline Reinforcement Learning
Federated learning aims to solve a global optimization problem by collectively learning from a group of clients without sharing data they possess. In offline reinforcement learning, an agent aims to learn an optimal policy for its behavior without accessing the environment. While federated …
-
Towards convergent offline reinforcement learning
Reinforcement learning (RL) is a leading method for automated sequential decision-making. However, RL is rarely used in the real world. This neglect is due to instability and inefficiency, often even occurring on benchmark problems. Conversely, the sweeping real-world impact of supervised learning …
-
Offline reinforcement learning via an optimization lens
… and frameworks to address challenges in offline reinforcement learning. More specifically, four perspectives have been studied to improve the efficiency and reliability of offline policy evaluation and learning: 1) We manage to improve the accuracy of off-policy policy evaluation, by …
-
Evaluation of sample efficiency in offline reinforcement learning
Reinforcement learning (RL) has shown great promise with algorithms learning in environments with large state and action spaces purely from scalar reward signals. A crucial challenge for current deep RL algorithms is that they require a tremendous amount of environment interactions for learning. …
-
Transformer-Based Offline Reinforcement Learning: Applications in Condition-Based Maintenance
… thesis bridges traditional optimization and deep learning by proposing a novel framework for maintenance decision-making in complex systems. We formulate the equipment replacement problem as a sequential decision-making problem that is suitable for reinforcement learning (RL). To address the …
-
Donsker-Varadhan policy matching for reward-free offline reinforcement learning
… studies an alternative method for reward-free offline reinforcement learning, investigating how well an agent can learn from a fixed dataset when reward guidance is missing, weak, or unreliable. A common way to handle this problem is behavior regularization, where the learned policy is kept …
-
Policy evaluation in offline reinforcement learning with low-rank structure
… the target policy, which is learned from an offline data that is generated from an unknown behavior policy, offline reinforcement learning, is crucial before deploying in the environment to mitigate the risk. Previous works provided the policy evaluation guarantees when all the state-action …
-
Leveraging Learned Models for Robust Decision Optimization and Offline Reinforcement Learning
Recent advances in machine learning (ML) have driven the adoption of predictive models for decision-making in data-rich environments in areas like natural language processing, computer vision, and autonomous systems. These models are valuable for decision optimization, predicting uncertain …
-
Offline Reinforcement Learning for Energy-Efficient Target Wake Time Scheduling in IEEE 802.11AX Networks
… on its own. This thesis investigates whether an offline reinforcement learning (RL) agent can learn a TWT scheduling policy that reduces device energy consumption while maintaining acceptable throughput and fairness across heterogeneous device classes. The approach formulates TWT scheduling as a …
-
OFFLINE REINFORCEMENT LEARNING FOR MDPS AND CONFOUNDED POMDPS: POLICY GRADIENT ALGORITHMS AND PESSIMISTIC MODEL-BASED APPROACHES
This dissertation aims to develop new offline reinforcement learning (RL) methods with theoretical guarantees. The first contribution is a model-based offline RL algorithm under the framework of Markov decision processes (MDPs). Existing approaches either emphasize theoretical aspects without …
-
LEARNING COMPLICATED HUMAN TASKS FOR ROBOTICS MANIPULATION WITH LIMITED DEMONSTRATIONS THROUGH ONLINE AND OFFLINE REINFORCEMENT LEARNING
… due to the inherent limitations of traditional reinforcement learning (RL) techniques. Among these are inefficient exploration, slow convergence, and high failure rates during the learning phases. This thesis addresses these challenges by introducing novel methodologies integrating human …
-
Towards safe and efficient offline reinforcement learning: learning safety constraints and expressive policies via generative modeling
Offline reinforcement learning (RL) presents a particularly challenging setting due to the lack of active environment interaction, making it critical to learn from static datasets in a way that both respects safety constraints and captures complex action distributions. This thesis investigates …
-
From Structure Learning to Decision-Making: Advanced Bayesian Clustering and Offline Reinforcement Learning Methods with Applications to Biomedical Data
… dissertation develops statistical and machine learning methods for extracting latent structure and learning decision policies from complex, high-dimensional healthcare data. Modern biomedical research generates rich observational data with high-dimensional covariates, multi-modal outcomes, and …
-
Group-Based Target Wake Time Scheduling to Improve Quality of Service in IEEE 802.11AX Networks using Offline Reinforcement Learning
… over time. This thesis investigates adaptive, offline reinforcement learning for group-based TWT scheduling. Although static TWT configurations can improve performance in stable workloads, they typically degrade when the load, queue occupancy, and contention intensity vary over time, leading to …
-
THREE ESSAYS ON DISCRETE CHOICE LATENT CLASS MODEL, ONLINE REUSABLE RESOURCE ALLOCATION AND TRANSFORMER MODEL FOR OFFLINE REINFORCEMENT LEARNING
… Re-evaluated Transformer model for model-free offline reinforcement learning tasks to both better learn state values and to address data coverage issues. The proposed model incorporates a novel and concise Re-evaluation mechanism in the sequential model, which leverages the trained model to …
-
Offline Reinforcement Learning for Optimizing and Evaluating Early Treatment Initiation Strategies in Treatment-Resistant Depression: A Retrospective Cohort Study Using Administrative Claims Data
… should be sequenced remains unknown. We applied offline reinforcement learning to administrative claims data to assess how early treatment strategies for TRD might be optimized. Methods. We conducted a retrospective cohort study of commercially insured U.S. adults with TRD, following individuals …
Page 1 of 3