TopRank: A practical algorithm for online stochastic ranking

Authors: Tor Lattimore, Branislav Kveton, Shuai Li, Csaba Szepesvari

NeurIPS 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental 6 Experiments We experiment with the Yandex dataset [15], a dataset of 167 million search queries. ... Top Rank is compared to Batch Rank [17] and Cascade KL-UCB [6]. ... The results are averaged over 10 runs.
Researcher Affiliation Collaboration Tor Lattimore Deep Mind Branislav Kveton Google Shuai Li The Chinese University of Hong Kong Csaba Szepesvári Deep Mind and University of Alberta
Pseudocode Yes Algorithm 1 Top Rank
Open Source Code No The paper mentions using 'Py Click' and the 'implementation of Batch Rank by Zoghi et al. [17]' but does not provide any links or explicit statements about releasing their own source code for Top Rank.
Open Datasets Yes We experiment with the Yandex dataset [15], a dataset of 167 million search queries. ... [15] Yandex. Yandex personalized web search challenge. https://www.kaggle.com/c/yandexpersonalized-web-search-challenge, 2013.
Dataset Splits No The paper mentions selecting '60 frequent search queries' and learning 'CMs and PBMs' but does not specify explicit training, validation, or test dataset splits (e.g., percentages or sample counts).
Hardware Specification No The paper does not provide specific details about the hardware used to run the experiments (e.g., GPU/CPU models, memory).
Software Dependencies No The paper mentions using 'Py Click' and an 'implementation of Batch Rank', but it does not provide specific version numbers for these or any other software dependencies.
Experiment Setup Yes Our goal is to rerank L = 10 most attractive items with the objective of maximizing the expected number of clicks at the first K = 5 positions. This is the same experimental setup as in Zoghi et al. [17]. ... The parameter δ in Top Rank is set as δ = 1/n, as suggested in Theorem 1.