Worst-case regret analysis of computationally budgeted online kernel selection

We study the problem of online kernel selection under computational constraints, where memory or time and prediction procedures is restricted to a fixed budget. In this paper, we analyze worst-case lower bounds on regret algorithm with subset observed examples, design algorithms enjoying corresponding upper bounds. also identify condition which constraints different from that constraints. To algorithms, reduce problems two sequential decision problems, is, expert advice multi-armed bandit an additional observation. Our invent some new techniques, such as sharing, hypothesis space discretization decoupled exploration-exploitation scheme. Numerical experiments regression classification are conducted verify our theoretical results.