New lower bounds established for quantum bandit algorithms

Researchers have established new minimax lower bounds for quantum multi-armed bandits (QMAB) and quantum linear bandits (QLB), improving upon prior work. The study proves lower bounds of $\Omega(K\log(T/K))$ for QMAB with $K$ arms over horizon $T$, and $\Omega(d\log(T/d))$ for $d$-dimensional QLB. These findings resolve questions about achieving regret independent of $T$ and reducing the dimension dependence in QLB. Complementing these theoretical results, an algorithm is presented for finite-action QLB that achieves linear dependence on $d$ for regret, an improvement over the previous $d^2$ complexity. AI

IMPACT Establishes theoretical limits for quantum bandit algorithms, potentially guiding future research in quantum machine learning optimization.

RANK_REASON The cluster contains an academic paper detailing new theoretical results and algorithms in quantum machine learning. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• Wan et al., 2023

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →