Accompanying code for the ICML 2023 submission titled 'Delayed Feedback in Kernel Bandits': https://dl.acm.org/doi/10.5555/3618408.3619855

We provide code for understanding the regret performance of BPE-Delay, GP-UCB-SDF, BPE or GP-UCB with varying levels of stochastic delayed feedback.

The GP-UCB class was used to understand the performance of GP-UCB-SDF when no delay is present.

In order to conduct experiments, follow these steps similar to those in the referenced repsoitory:

• Generate an objective function as described in the referenced repository.
• Run main.py to execute BPE, BPE-Delay, GP-UCB-SDF or GP-UCB.

python main.py [function_name] [beta] [poisson_parameter] [algorithm]

• The supported choice of beta is [6]. Edit the dictionary BETA at the top of main.py to add more choices of beta.