Fatigue-Aware Bandits for Dependent Click Models

Authors: Junyu Cao, Wei Sun, Zuo-Jun (Max) Shen, Markus Ettl3341-3348

AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In this section, we perform four sets of numerical experiments to evaluate the performance of our online learning algorithms.
Researcher Affiliation Collaboration 1University of California, Berkeley, California 94720 2IBM Research, Yorktown Height, New York 10591
Pseudocode Yes Algorithm 1: Determine the optimal sequence S to the offline combinatorial problem
Open Source Code No The paper does not provide any statement or link indicating that source code for the described methodology is publicly available.
Open Datasets Yes The data is from Taobao, which contains 26 million ad display and click logs from 1,140,000 randomly sampled users from the website of Taobao for 8 days (5/6/2017-5/13/2017). https://tianchi.aliyun.com/datalab/data Set.html?spm=5176. 100073.0.0.14d53ea7Rleuc9&dataId=56
Dataset Splits No The paper mentions using data from Taobao but does not provide specific details on how it was split into training, validation, or test sets.
Hardware Specification No The paper does not provide any specific hardware details (e.g., GPU/CPU models, memory) used for running its experiments.
Software Dependencies No The paper does not provide specific ancillary software details with version numbers.
Experiment Setup Yes We consider a setting with three categories, and each contains 10 items. Items intrinsic relevance u is uniformly generated from [0, 0.5]. The known discount factor is set to f(r) = exp( 0.1 (r 1)). We fix the resuming probability after non-clicks to q = 0.7, and compare three cases by varying the resuming probabilities after clicks, i.e, Case 1: g = 0.95; Case 2: g = 0.85; Case 3: g = 0.75. For each case, we run 20 independent simulations.