Representation transfer for sequential multitask linear bandits

This repository is the official implementation of "Beyond Task Diversity: Provable Representation Transfer for Sequential Multi-Task Linear Bandits" and "Efficient Algorithms for Lifelong Learning Linear Bandits"

Requirements

Python 3.10

git clone https://github.com/duongnhatthang/BOSS.git
cd BOSS
pip install -r requirements.txt

Evaluation

Interactive Notebook

Results

Our algorithm achieves the following performance on a synthetic dataset:

Regret over tasks	Estimation error of B	Estimation error of $\theta_n$

Estimation error of the angle between $(\theta_n, \hat{\theta}_n)$	Regret over T, average across all tasks

Contributing

Apache 2.0

Note

The code is referencing this repo.