ConceptMaximum Information Gain / γ_T / gamma_T
Maximum Information Gain
A key quantity in kernelized bandit theory that characterizes the maximum mutual information an algorithm can obtain about an unknown function through T rounds of interaction. It determines the hardness boundary of kernelized bandit problems and governs the theoretical complexity of the learning task.
Timeline (last 90 days)
Oct 9
The maximum information gain γ_T characterizes the upper limit on the information about the unknown function that an algorithm can obtain through observations in T rounds of interaction, and largely determines the difficulty boundary of the kernelized bandit problem.
Unverified50%