Impact of Representation Learning in Linear Bandits

Authors: Jiaqi Yang, Wei Hu, Jason D. Lee, Simon Shaolei Du

ICLR 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We also present experiments on synthetic and realworld data to illustrate our theoretical findings and demonstrate the effectiveness of our proposed algorithms.
Researcher Affiliation Academia Jiaqi Yang Tsinghua University yangjq17@gmail.com Wei Hu Princeton University huwei@cs.princeton.edu Jason D. Lee Princeton University jasonlee@princeton.edu Simon S. Du University of Washington ssdu@cs.washington.edu
Pseudocode Yes Algorithm 1: MLin Greedy: Multi-task Linear Bandit with Finite Actions; Algorithm 2: E2TC: Explore-Explore-Then-Commit
Open Source Code No The paper does not contain any explicit statements or links indicating that the source code for the described methodology is publicly available.
Open Datasets Yes We create a linear bandits problem on MNIST data (Le Cun et al., 2010)
Dataset Splits No The paper mentions 'N = 10000' total rounds but does not specify train, validation, or test dataset splits for the experiments.
Hardware Specification No The paper does not provide any specific details about the hardware used to run the experiments.
Software Dependencies No The paper does not provide any specific software names with version numbers, nor any self-contained solvers or specialized packages with versions.
Experiment Setup Yes We fix K = 5 and N = 10000 for all simulations on finite-action setting. We vary k, d and T to compare Algorithm 1 and the naive algorithm. We emphasize that the y-axis in our figures corresponds to the regret per task, which is defined as RN,T /T. We fix K = 5, N = 10000. We create a linear bandits problem on MNIST data (Le Cun et al., 2010) to illustrate the effectiveness of our algorithm on real-world data. We fix K = 2 and create T = 10 2 tasks and each task is parameterized by a pair (i, j), where 0 i < j 9. We consider k = 2, 3 in our experiments. The noise εn,t N(0, 1) are i.i.d. Gaussian random variables. To verify our theoretical results, we consider a hyper-parameter c {0.5, 1, 1.5, 2}. For each c, we run E2TC with N1 = dck q T and N2 = k N.