Near-Optimal MNL Bandits Under Risk Criteria

Authors: Guangyu Xi, Chao Tao, Yuan Zhou10397-10404

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

Reproducibility Variable Result LLM Response
Research Type Experimental As a complement, we also conduct experiments with both synthetic and real data to show the empirical performance of our proposed algorithms.
Researcher Affiliation Academia Guangyu Xi, 1 Chao Tao, 2 Yuan Zhou 3 1 University of Maryland, College Park 2 Indiana University Bloomington 3 University of Illinois at Urbana-Champaign
Pseudocode Yes Algorithm 1: Risk Aware UCB(N, K, r, U)
Open Source Code Yes 1Please refer to https://github.com/Alanthink/aaai2021 for the source code.
Open Datasets Yes In this experiment, we consider the UCI Car Evaluation Database dataset from the Machine Learning Repository (Dua and Graff 2017)
Dataset Splits No The paper describes using synthetic and real datasets, but it does not specify explicit training, validation, or test splits, nor does it mention cross-validation. It only states the number of repetitions for experiments.
Hardware Specification No The paper does not provide any specific details about the hardware (e.g., CPU, GPU, memory) used to run the experiments.
Software Dependencies No The paper mentions that algorithms are "implemented in Python3" and use the "Bandit Py Lib library," but it does not provide specific version numbers for Python, the library, or any other software dependencies.
Experiment Setup Yes In this experiment, we fix the number of products N = 10, cardinality limit K = 4, horizon T = 10^6, and set the goal to be U = CVa R0.5. We generate 10 uniformly distributed random input instances where vi [0, 1] and ri [0.1, 1]. For each input instance, we run 20 repetitions and compute their average as the regret.