This library corresponds to the Online Limited Memory Neural-Linear Bandits with Likelihood Matching paper, ICML 2021.
The code is based on the "Deep Bayesian Bandits Showdown: An Empirical Comparison of Bayesian Deep Networks for Thompson Sampling" github repository.
All the datasets can be found in UCI Machine Learning Repository, and should be placed under contextual_bandtis/datasets folder.
Run at terminal:
python3 main.py
At the main.py, set the method into:
- neural-linear-lm (Our method)
- neural-linear (full memory NeuralTS)
- linear (LinearTS)
- neural-linear-ntk (NTK version of limited memory NeuralTS)
At dataset sepcify the wanted dataset (an unknown dataset will cause an error). For amazon dataset, LinearTS do not work.
- tensorflow-gpu 1.15
- absl-py 0.11
- scipy 1.5.4