Anfonwch hwn fel neges destun: Gradient-based stochastic optimization methods in bayesian experimental design