Regret Bounds for Safe Gaussian Process Bandit Optimization | Read Paper on Bytez