New perspectives on online prediction and decision making
Abstract/Contents
- Abstract
- The problem of predicting the future has a long history, and it has become ever more important in recent years, as uncertainties arise with evolving geopolitical crises and quickly unfolding pandemics around the globe. Can we accurately predict the future and make near-optimal decisions, even if the world is complex, behaves arbitrarily, and may not satisfy any of the strong traditional assumptions in the literature? This thesis studies several different settings of sequential prediction and decision-making, especially when distributional assumptions are absent: 1. Sequential calibration: Facing a sequence of intrinsically noisy outcomes, to what extent can probabilistic forecasts be well-calibrated? We develop a novel "sidestepping" technique to prove the first non-trivial lower bound on the calibration error in sequential binary prediction. 2. Selective prediction of arbitrary data: Can we accurately predict the unseen data in the future, even when distributional structure of data and advice from domain experts are absent? We formulate models of selective prediction and learning and study the optimal accuracy of a forecaster that may choose to predict over a carefully selected interval. 3. Optimal stopping with costly probing: Can we make high-stake decisions optimally even when obtaining information is difficult and associated with high opportunity costs? We define and study the problem of online pen testing, which abstracts the challenge in this decision-making problem.
Description
| Type of resource | text |
|---|---|
| Form | electronic resource; remote; computer; online resource |
| Extent | 1 online resource. |
| Place | California |
| Place | [Stanford, California] |
| Publisher | [Stanford University] |
| Copyright date | 2023; ©2023 |
| Publication date | 2023; 2023 |
| Issuance | monographic |
| Language | English |
Creators/Contributors
| Author | Qiao, Mingda |
|---|---|
| Degree supervisor | Valiant, Gregory |
| Thesis advisor | Valiant, Gregory |
| Thesis advisor | Charikar, Moses |
| Thesis advisor | Tan, Li-Yang |
| Degree committee member | Charikar, Moses |
| Degree committee member | Tan, Li-Yang |
| Associated with | Stanford University, School of Engineering |
| Associated with | Stanford University, Computer Science Department |
Subjects
| Genre | Theses |
|---|---|
| Genre | Text |
Bibliographic information
| Statement of responsibility | Mingda Qiao. |
|---|---|
| Note | Submitted to the Computer Science Department. |
| Thesis | Thesis Ph.D. Stanford University 2023. |
| Location | https://purl.stanford.edu/gk910mk7857 |
Access conditions
- Copyright
- © 2023 by Mingda Qiao
- 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...