Researchers have developed a novel data structure designed for efficient online proportional sampling in high-dimensional domains. This structure addresses the challenge of managing complex, dynamically evolving weight functions and partitions, which can grow exponentially in complexity. The proposed solution offers provable bounds on data structure depth, achieving $O(\sqrt{\sigma T})$ under a $\sigma$-smoothed adversary and $O(\log T)$ under a random-order adversary. These advancements enable efficient no-regret algorithms for online learning with piecewise-structured rewards, providing sublinear regret guarantees under both full-information and bandit feedback. AI
IMPACT Enables more efficient algorithms for online learning and parameter tuning in complex, dynamic environments.
RANK_REASON The item is an academic paper submitted to arXiv detailing a new data structure and algorithms for online proportional sampling. [lever_c_demoted from research: ic=1 ai=1.0]
- alphaXiv
- Amirmahdi Mirfakhar
- arXiv
- CatalyzeX Code Finder for Papers
- CORE Recommender
- cs.AI
- DagsHub
- Gotit.pub
- Hugging Face
- IArxiv Recommender
- Influence Flower
- machine learning
- ScienceCast
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
