摘要: 过拟合 L1/L2 正则化 正则化是w权重的值偏小,趋势每层的输出值偏小。对于激活函数来说类似于把输出聚集在0附近, 而在0附近激活函数类似线性,这就降低了激活函数的非线性功能 原因:通过正则化,使神经元的输出偏小,类似于消除了部分神经元,使得网络变得简单。存在一个状态使得模型比较好的拟合输入数据。 阅读全文
posted @ 2020-10-19 10:17 ab229693 阅读(168) 评论(0) 推荐(0)
摘要: torch.nn.utils.rnn: pack_padded_sequence() pad_packed_sequence() Notice: The padded embedding metrix must be sorted by the ground length of each sente 阅读全文
posted @ 2019-03-28 10:53 ab229693 阅读(98) 评论(0) 推荐(0)
摘要: Open domain QA Overview The whole system is consisted with Document Retriever and Document Reader. The Document Retriever returns top five Wikipedia a 阅读全文
posted @ 2019-02-24 08:14 ab229693 阅读(236) 评论(0) 推荐(0)
摘要: Week Six F Score $$\begin{aligned} P &= &\dfrac{2}{\dfrac{1}{P}+\dfrac{1}{R}}\\ &= &2 \dfrac{PR}{P+R} \end{aligned}$$ Week Seven Support Vector Machin 阅读全文
posted @ 2018-12-03 10:37 ab229693 阅读(68) 评论(0) 推荐(0)