Delay as Payoff in MAB

Authors: Ofir Schlisselberg, Ido Cohen, Tal Lancewicki, Yishay Mansour

AAAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Finally, we accompany our theoretical results with an empirical evaluation. We conducted synthetic experiments for both the cost and reward settings, using the algorithms in Table 1 as baselines. We show results on two representative distributions: Truncated Normal (bounded in [0, D]) and Bernoulli. ... Figure 1 shows the average cumulative regret over 10 runs.
Researcher Affiliation Collaboration 1Tel Aviv University 2Google Research EMAIL, EMAIL
Pseudocode Yes Algorithm 1 Protocol1 ... Algorithm 2 Cost Successive Elimination (CSE) ... Algorithm 3 Bounded Doubling Successive Elimination ... Algorithm 4 Reward Successive Elimination ... Algorithm 5 Bounded Halving Successive Elimination (BHSE)
Open Source Code No The paper does not contain an explicit statement about releasing code for the described methodology, nor does it provide a link to a code repository.
Open Datasets No We conducted synthetic experiments for both the cost and reward settings... For the truncated Normal we sample K means and standard deviations (std)... For the Bernoulli distribution, we sample K probabilities pi uniformly in [0, 1]...
Dataset Splits No The paper states parameters for synthetic data generation (e.g., T=150,000, K=30, D=5000) but does not provide specific training/test/validation dataset splits, cross-validation, or other data partitioning details.
Hardware Specification No The paper does not provide any specific hardware details such as GPU/CPU models, processor types, or memory used for running the experiments.
Software Dependencies No The paper does not provide specific version numbers for any software components, libraries, or programming languages used in the experiments.
Experiment Setup Yes All experiments use T=150, 000, K=30 and D=5000. For the truncated Normal we sample K means and standard deviations (std), and adjust them to get a truncated version... For the Bernoulli distribution, we sample K probabilities pi uniformly in [0, 1]... Figure 1 shows the average cumulative regret over 10 runs.