之前学习到L1,L2 Regularization(正则化)和各种机器学习算法的时候,大家也许都见过这一张图。
并且书本会告诉你,Lasso tends to generate sparser solutions than a quadratic regularizer (ridge regression)
但是你们具体想过为啥L1会更大几率稀疏么?下面推荐一个PPT Slide 帮助你们更好地理解正则化Regularization 。
感谢Darren Wang, 原文链接:https://www.slideshare.net/darrenyowbangwang?utm_campaign=profiletracking&utm_medium=sssite&utm_source=ssslideview
Leave A Comment