Multi-armed bandits with side information
Abstract/Contents
- Abstract
- Development of personalized strategies has attracted much attention in recent years. Advances in Information Technology has led to the inundation of information, which has provided the impetus for the development of personalized strategies in diverse fields such as medicine, marketing, finance, and education. A methodological framework for the development of these personalized strategies is the theory of multi-armed bandits with side information, also called "contextual" bandits theory, which is an extension of classical (i.e., context-free) multi-armed bandits. This thesis represents an attempt to develop this theory To begin with, we consider the design of clinical trials for developing and testing biomarker-guided personalized therapies. Biomarker-guided personalized therapies offer great promise to improve drug development and patient care, but also pose difficult challenges in designing clinical trials for the development and validation of these therapies. After a review of the existing approaches, we describe new adaptive designs to address these challenges, first for clinical trials in new drug development and then for comparative effectiveness trials involving approved treatments. With the insight from the real-world application, we next consider general multi-armed bandits with side information, and categorizes the approaches to address the multi-armed bandit problems with side information into two groups. One approach uses parametric regression model and the other uses nonparametric regression for the relationship between the reward function of the side information for each arm. We review the existing literature and describe new treatment allocation policies inspired by the design of the biomarker-guided personalized therapy problem. We show that the new treatment allocation policies are applicable to a much more general class of problems than the policies proposed in the literature, and show the improvement of the new policies in simulation studies.
Description
Type of resource | text |
---|---|
Form | electronic; electronic resource; remote |
Extent | 1 online resource. |
Publication date | 2014 |
Issuance | monographic |
Language | English |
Creators/Contributors
Associated with | Kim, Dong Woo |
---|---|
Associated with | Stanford University, Department of Electrical Engineering. |
Primary advisor | Lai, T. L |
Thesis advisor | Lai, T. L |
Thesis advisor | Boyd, Stephen P |
Thesis advisor | Van Roy, Benjamin |
Advisor | Boyd, Stephen P |
Advisor | Van Roy, Benjamin |
Subjects
Genre | Theses |
---|
Bibliographic information
Statement of responsibility | Dong Woo Kim. |
---|---|
Note | Submitted to the Department of Electrical Engineering. |
Thesis | Thesis (Ph.D.)--Stanford University, 2014. |
Location | electronic resource |
Access conditions
- Copyright
- © 2014 by Dong Woo Kim
- License
- This work is licensed under a Creative Commons Attribution Non Commercial 3.0 Unported license (CC BY-NC).
Also listed in
Loading usage metrics...