Backpropagation - Wikipedia
7
公开标注数
6
参与人数
2026-06-18 09:09:18
首次 Whisper
本页的公开 Whisper
划选高亮2026-07-26 16:05:42
原文高亮摘录
“Strictly speaking, the term backpropagation refers only to an algorithm for efficiently computing the gradient, not how the gradient is used”
Whisper 随想笔记
So it's just the gradient part, like a tool, not the whole learning thing.
划选高亮2026-07-26 13:02:42
原文高亮摘录
“efficient application of the chain rule to neural networks”
Whisper 随想笔记
So many tutorials skip the 'efficient' part and just drown you in math.
划选高亮2026-07-26 12:53:42
原文高亮摘录
“efficient application of the chain rule to neural networks”
Whisper 随想笔记
Wait, isn't that just the chain rule with extra steps?
划选高亮2026-07-26 09:59:42
原文高亮摘录
“backpropagation is a gradient computation method commonly used for training a neural network”
Whisper 随想笔记
basically the chain rule on steroids, right?
划选高亮2026-07-26 09:50:42
原文高亮摘录
“backpropagation is a gradient computation method commonly used for training a neural network”
Whisper 随想笔记
i always mix this up with the optimizer, makes sense now
划选高亮2026-07-26 09:41:42
原文高亮摘录
“backpropagation is a gradient computation method commonly used for training a neural network”
Whisper 随想笔记
so it's just gradients, not the whole algorithm, huh
划选高亮2026-06-18 09:09:18
原文高亮摘录
“the term is often used loosely to refer to the entire learning algorithm”
Whisper 随想笔记
Good point, most people mean the whole training loop when they say backprop.
分享本页 Whisper
短链接
https://domwhisper.com/s/3b380be231de嵌入代码
<iframe src="https://domwhisper.com/embed/3b380be231de" width="100%" height="480" style="border:0;border-radius:16px" loading="lazy"></iframe>