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 29 for “"Reward design"”.
-
Reward Design with Program Graphs for Reinforcement Learning Guided Training of Large Language Models for Program Synthesis
… up brings an additional challenge: the task of designing a reward function for the Code LLM. This thesis looks at reward design within the context of RL for Code LLMs. We hypothesize that a better reward signal for Code LLMs undergoing RL-based training will result in a better resulting model, …
-
Robust and Flexible Reward Modeling for LLM Alignment
… promising ways to align language models is by designing a reward function that measures alignment with human values, and training the language model to maximize this reward. In this thesis, we focus on two approaches reward design: reward design from external feedback signals and reward …
-
ROBUST AND AGILE QUADRUPED CONTROL VIA REWARD-SHAPING REINFORCEMENT LEARNING
… (RL) has emerged as a prominent approach, and reward-shaping is essential in robot training. To address this, our work introduces two unified reward design frameworks, each tailored to distinct task categories. In the first part, we present a teacher–student RL architecture for fault-tolerant …
-
ESSAYS ON AUCTIONS AND CONTESTS
… contests. The first chapter studies the optimal reward design in contest setting with complete information; the second chapter studies the design of disclosure policy in first price auctions; the third chapter examines the design of an efficient auction in a bi-dimensional private information …
-
Learning versatile skills with sample-efficient reinforcement learning
… The scope of usable data is then extended to reward-free, mixed-quality, and multi-embodiment offline datasets by leveraging a task-agnostic dynamics model learned from such data, yielding high-performance policies with improved sample efficiency across 72 visual motor control tasks. (3) …
-
Optimization and imitation for pandemic policy and financial decision making
… to reproduce trading behaviours, without manual reward design needed in reinforcement learning. In chapter 2, we review main imitation learning techniques and applications in finance. Analyses provide error bounds and generalisation guarantees.</p> <p>The final chapter draws conclusions and list …
-
Imitating optimized trajectories for dynamic quadruped behaviors
… solutions without the need for meticulous reward design. While much work uses motion capture data or hand-crafted trajectories as the reference motion, relatively little work has explored the use of reference motions coming from model-based trajectory optimization. This may be advantageous …
-
The Effects of Reward Nature and Reward Contingency on Employee Voluntary Training
… examine how two attributes of performance-based rewards – reward nature (hedonic vs. utilitarian) and reward contingency (goal-based vs. piece-rate) – affect employee total effort, voluntary training effort proportion, and performance. Using a lab experiment, I find that the effect of hedonic …
-
AN EVALUATION OF REINFORCEMENT LEARNING ALGORITHMS IN VIDEO GAME DEVELOPMENT
… remains limited due to computational overhead, reward-design challenges, and unpredictable AI behaviors. This thesis investigates how RL algorithms—specifically Advantage Actor-Critic (A2C), Deep Q-Network (DQN), and Proximal Policy Optimization (PPO)—can be applied to three different genres of …
-
Online Active Preference Learning from Sparse Expert Feedback in Continuous Control
… reinforcement learning, direct reward design can bedifficult when desired behavior is multi-dimensional, and expert feedback is often too limited to query exhaustively. This thesis develops an online active-preference learning framework that combines a fixed expert log-density …
-
Deep-Reinforcement Learning Agents in Autonomous Directional Drilling and Geosteering
… optimization (PPO), we demonstrated that careful reward design and hyperparameter tuning significantly improve agent efficiency, reducing training time and enabling robust policy learning in uncertain drilling environments. Domain randomization further enhanced generalization, allowing agents to …
-
Deep-Reinforcement Learning Agents in Autonomous Directional Drilling and Geosteering
… optimization (PPO), we demonstrated that careful reward design and hyperparameter tuning significantly improve agent efficiency, reducing training time and enabling robust policy learning in uncertain drilling environments. Domain randomization further enhanced generalization, allowing agents to …
-
Deep-Reinforcement Learning Agents in Autonomous Directional Drilling and Geosteering
… optimization (PPO), we demonstrated that careful reward design and hyperparameter tuning significantly improve agent efficiency, reducing training time and enabling robust policy learning in uncertain drilling environments. Domain randomization further enhanced generalization, allowing agents to …
-
Deep-Reinforcement Learning Agents in Autonomous Directional Drilling and Geosteering
… optimization (PPO), we demonstrated that careful reward design and hyperparameter tuning significantly improve agent efficiency, reducing training time and enabling robust policy learning in uncertain drilling environments. Domain randomization further enhanced generalization, allowing agents to …
-
Using natural language to aid task specification in sequential decision making problems
… specifying the task to the learning agent, i.e. designing effective reward functions for reinforcement learning and providing demonstrations for imitation learning, are often cumbersome and time-consuming. Further, designing reward functions and providing a set of demonstrations that sufficiently …
-
PHYSICS-INFORMED REINFORCEMENT LEARNING FOR SEQUENTIAL BASE-PAIRING AND RNA SECONDARY-STRUCTURE PREDICTION
… has received less attention than inverse-design RL, where the agent designs a sequence for a target structure. The legality-constrained environment supports pair, skip, and stop actions under Watson–Crick and GU pairing rules, minimum loop lengths, and singleoccupancy constraints. The …
-
Machine learning for optimization and decision support in complex system-of-systems: applications in microelectronics, healthcare, smart cities, and the circular economy
… safety systems via policy learning and augmented reward design. Simulations demonstrate improvements in coordination, efficiency, and system responsiveness. In circular economy digital marketplaces, time-series forecasting (e.g., AutoRegressive Integrated Moving Average (ARIMA), LSTM), Gradient …
-
Deep-EC2: A Non-Myopic Active Feature Acquisition Framework
… rely on reinforcement learning (RL), with its reward design, rollouts, and hyperparameter sweeps. This thesis proposes Deep-EC2, an AFA framework that replaces reinforcement learning with supervised regression onto analytical targets derived from the Equivalence Class Edge Cutting (EC2) …
-
Data Driven Turbine Control in Tidal Power Generation: Development and comparison of data driven turbine control algorithms in a simulated underwater kite environment
… the surrounding water-flow. To achieve this, designing a generator control signal that applies optimal torque on the turbine shaft is essential. This project aims to develop data-driven control systems for the kites generator that increases the power output relative to a baseline. ODE-based …
-
Phase-aware 3D safety planning and recovery for collaborative robot transport under dynamic human interference
… pipelines require extensive training, careful reward design, and more difficult debugging in exactly the part of the system where transparent safety assurances are most important. For that reason, reinforcement learning is reserved for future high-level route preference rather than being placed …
Page 1 of 2