Stochastic Gradient Descent Convergence:
MapReduce:
Large Scale Machine Learning:
b) SGD 適合大規模數據集,因爲batch gradient在大數據集下,每次計算太慢
c)並不能保證每次下降,SGD每次是隨機的,可能上升
Stochastic Gradient Descent Convergence:
MapReduce:
Large Scale Machine Learning:
b) SGD 適合大規模數據集,因爲batch gradient在大數據集下,每次計算太慢
c)並不能保證每次下降,SGD每次是隨機的,可能上升