引用stochastic_monk · 2017-10
“Deep Forest [Deep Forest is a method which claims to work well, but it's somewhat odd; It's not quite what I would have imagined.”
评价 Deep Forest 号称效果好,但方法有些古怪
用决策树集成逐层堆叠,不用反向传播也能搭深度模型
探讨用不可微模块构建深度模型的可能性,推测深度神经网络成功源于三个特征:逐层处理、模型内特征变换、足够的模型复杂度。作者据此提出 gcForest 方法,生成具备这些特征的“深度森林”,这是一种决策树集成方法,超参数比深度神经网络少得多,模型复杂度可依数据自动确定。实验表明其对超参数设置相当稳健,多数情况下即使跨领域数据也能用同一默认设置取得优秀表现。
评论者称它是声称效果很好的方法,但有点怪、不是自己设想的样子,后来也被当作可以训练深度决策树森林的例子。
“Deep Forest [Deep Forest is a method which claims to work well, but it's somewhat odd; It's not quite what I would have imagined.”
“It is possible to train deep decision trees forests [1].”