We used stochastic gradient ascent to maximize the log-likelihood.
我们使用随机梯度上升来最大化对数似然。
In large-scale logistic regression, stochastic gradient ascent updates the parameters using a noisy gradient estimate from each mini-batch, which often speeds up training despite added variance.
在大规模逻辑回归中,随机梯度上升用每个小批量数据得到的带噪声梯度估计来更新参数;尽管方差更大,但常常能加快训练。