High-Dimensional Prediction for Sequential Decision Making
Authors: Georgy Noarov, Ramya Ramalingam, Aaron Roth, Stephan Xie
ICML 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Theoretical | Our central contribution is a general and efficient algorithmic framework for event-unbiased prediction of vector quantities in the online adversarial setting. It accommodates arbitrary finite collections of (unweighted or weighted) conditioning events, which may depend on external contexts or on the predictions themselves, and guarantees optimally converging bias conditional on each of these events. We introduce the framework, and formally derive the algorithm and its guarantees, in Section 2. |
| Researcher Affiliation | Academia | 1Department of Computer and Information Science, University of Pennsylvania 2Machine Learning Department, Carnegie Mellon University. Correspondence to: Georgy Noarov <EMAIL>. |
| Pseudocode | Yes | Algorithm 1 Unbiased Prediction Initialize ˆs0 = 0d, s0 = 0d, E0( ) 0n, and AMs Mw C. for t = 1 . . . T do Get context xt X, and define Et( ) := E(xt, ). Get new weights wt by updating AMs Mw C with losses: ℓouter t 1 σ Et 1 j (ˆst 1) (st 1,i ˆst 1,i) σ,i,j Initialize a new instance of AS and any stent 0 S. for τ = 1 . . . t2 do Get simulated prediction stent τ by updating AS with: j [n] wt,(σ,i,j) Et j stent τ 1 ! end for Set st Unif {stent 0 , stent 1 , . . . , stent t2 } . Predict ˆst st. Observe true state st. end for |
| Open Source Code | No | The paper describes theoretical algorithms and frameworks, but does not provide any statement about concrete access to source code, a repository link, or code in supplementary materials. |
| Open Datasets | No | The paper presents a theoretical framework and algorithms without conducting experiments on specific datasets. It does not provide any concrete access information for publicly available datasets. |
| Dataset Splits | No | The paper is theoretical and does not involve empirical evaluation on datasets, thus no dataset split information is provided. |
| Hardware Specification | No | The paper describes theoretical algorithms and frameworks. It does not provide specific hardware details for running experiments, as no experiments are reported. |
| Software Dependencies | No | The paper is theoretical, focusing on algorithmic frameworks and proofs. It does not provide specific ancillary software details with version numbers. |
| Experiment Setup | No | The paper is theoretical and does not present experimental results. Therefore, it does not contain specific experimental setup details, hyperparameters, or training configurations. |