OptimalOrderSimpleRegretforGaussianProcess Bandits

Neural Information Processing Systems 

Our sequential optimization setting is as follows. Consider an objective functionf defined over a domainX Rd, whered N is the dimension of the input.