https://developers.google.cn/machine-learning/crash-course/regularization-for-sparsity/l1-regularization

L₁ 正则化

减少所有权重的绝对值 - 证明对宽度模型非常有效

TIM截图20180401125731.jpg