Accelerating Monte Carlo Bayesian Inference via Approximating Predictive Uncertainty over Simplex