Comments on: [转载][翻译] 卷积神经网络的直观解释

By: robber run

robber run — Tue, 15 Sep 2026 07:20:23 +0000

This translation actually makes the backprop diagram way clearer than the original English version. The way they explain the gradient flow in the pooling layer hits differently than I expected.

By: product copy generator

product copy generator — Mon, 14 Sep 2026 09:04:50 +0000

终于有人把池化层说清楚了，不用对着公式死磕。

By: plinko clicker game

plinko clicker game — Sun, 13 Sep 2026 19:45:47 +0000

The sliding filter explanation is the part that actually clicked for me, especially seeing how the 3x3 kernel slides over the 5x5 matrix to generate that activation map. I still get why LeNet is such a good starting point though, even if the math part feels a bit dry.