Thresholding Bandits with Augmented UCB

Authors: Subhojyoti Mukherjee, Naveen Kolar Purushothama, Nandan Sudarsanam, Balaraman Ravindran

IJCAI 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We conduct extensive simulation experiments to validate the performance of Aug UCB.
Researcher Affiliation Academia Subhojyoti Mukherjee1,Naveen Kolar Purushothama2,Nandan Sudarsanam3,Balaraman Ravindran4 1,4Department of Computer Science & Engineering, Indian Institute of Technology Madras 2Department of Electrical Engineering, Indian Institute of Technology Madras 3Department of Management Studies, Indian Institute of Technology Madras
Pseudocode Yes Algorithm 1 Aug UCB
Open Source Code No The paper does not provide any statement or link indicating the availability of open-source code for the described methodology.
Open Datasets No The paper describes simulation experiments where reward distributions are generated (e.g., 'Gaussian with means r1:4 = 0.2 + (0 : 3) 0.05'), but does not use or provide a publicly available dataset with concrete access information.
Dataset Splits No The paper describes simulation experiments for a multi-armed bandit problem, which does not involve traditional train/validation/test dataset splits. Performance is evaluated by tracking error percentage over time in repeated runs.
Hardware Specification No The paper describes simulation experiments but does not provide any specific details about the hardware (e.g., CPU, GPU models, memory) used to run these simulations.
Software Dependencies No The paper does not list specific software dependencies with version numbers (e.g., programming languages, libraries, or frameworks with their versions) used for the experiments.
Experiment Setup Yes Across all experiments consists of K = 100 arms (indexed i = 1, 2, , 100) of which Sτ = {6, 7, , 10}, where we have fixed τ = 0.5. In all the experiments, each algorithm is run independently for 10000 time-steps. At every time-step, the output set, ˆSτ, suggested by each algorithm is recorded; the output is counted as an error if ˆSτ = Sτ. In Figure 1, for each experiment, we have reported the percentage of error incurred by the different algorithms as a function of time; Error percentage is obtained by repeating each experiment independently for 500 iterations, and then respectively computing the fraction of errors. The details of the considered experiments are as follows. Experiment-1: The reward distributions are Gaussian with means r1:4 = 0.2 + (0 : 3) 0.05, r5 = 0.45, r6 = 0.55, r7:10 = 0.65 + (0 : 3) 0.05 and r11:100 = 0.4. The corresponding variances are σ2 1:5 = 0.5 and σ2 6:10 = 0.6, while σ2 11:100 is chosen independently and uniform in the interval [0.38, 0.42].