However, when the variance of the noisy gradient is large, the algorithm might spend much time bouncing around, leading to slower convergence and worse ...
確定! 回上一頁