Please enable JavaScript.
Coggle requires JavaScript to display documents.
模型可解釋性 - Coggle Diagram
模型可解釋性
-
建模之後
隱層分析方法
Visualizing and Understanding Convolutional Networks - 反卷积 (Deconvolution),Saliency Map(Guided Back Propagation),Class Activation Map,Mask。。。
Network Dissection: Quantifying Interpretability of Deep Visual Representations- 實驗發現:随着层数的增加,神经网络学到的概念类型也逐渐变得高级。
-
-
-
-