結果 : gradient based stochastic optimization methods in bayesian experimental design