GitHub - hetankevin/probono: Algorithms for accelerating personalization (in the context of linear bandit recommendations) to new users, given access to embeddings from other users with heterogenous tastes and preferences. Novel algorithm for learning subspaces, which we exploit via a modified LinUCB algorithm. Accepted to ICML 2025.

Branches Tags

Name		Name	Last commit message	Last commit date
Latest commit History 9 Commits
archive		archive
data		data
figs		figs
probono		probono
matplotlibrc		matplotlibrc
movielens.ipynb		movielens.ipynb
pmf.ipynb		pmf.ipynb
pmfSmall.ipynb		pmfSmall.ipynb
simulation.ipynb		simulation.ipynb
sold.ipynb		sold.ipynb
sold_movielens.ipynb		sold_movielens.ipynb
sold_movielens_pmf.ipynb		sold_movielens_pmf.ipynb
sold_sample_complexity_movielens.ipynb		sold_sample_complexity_movielens.ipynb

About

Algorithms for accelerating personalization (in the context of linear bandit recommendations) to new users, given access to embeddings from other users with heterogenous tastes and preferences. Novel algorithm for learning subspaces, which we exploit via a modified LinUCB algorithm. Accepted to ICML 2025.

openreview.net/forum?id=LjmXenyVeQ

bandits embeddings linear-bandits personalization recommender-system