摘要: 简单的说,Label Smoothing就是把one-hot向量从[0,0,1,0,0,0,...,0]变成[0.01,0.01,0.8,0.01,0.01,0.01,...,0.01],用公式表示,就是 其中,k是类别数量,a是一个较小的数.这样做的目的是为了缓解模型过于武断的问题,增强模型的泛化 阅读全文
posted @ 2024-07-29 21:14 MSTK 阅读(11) 评论(0) 推荐(0) 编辑