Regularization (mathematics) - Wikipedia
5
公开标注数
5
参与人数
2026-07-21 09:39:47
首次 Whisper
本页的公开 Whisper
划选高亮2026-07-21 13:00:47
原文高亮摘录
“Stops training when validation performance deteriorates, preventing overfitting by halting before the model memorizes training data.”
Whisper 随想笔记
But what if validation noise makes you stop too early?
划选高亮2026-07-21 12:51:47
原文高亮摘录
“Stops training when validation performance deteriorates, preventing overfitting by halting before the model memorizes training data.”
Whisper 随想笔记
So basically it's like quitting while you're ahead, got it.
划选高亮2026-07-21 09:57:47
原文高亮摘录
“regularization is a process that converts the answer to a problem to a simpler one.”
Whisper 随想笔记
That's how I got my neural net to stop memorizing the training data, ha.
划选高亮2026-07-21 09:48:47
原文高亮摘录
“regularization is a process that converts the answer to a problem to a simpler one.”
Whisper 随想笔记
But doesn't 'simpler' sometimes mean less accurate? I've seen that fail in practice.
划选高亮2026-07-21 09:39:47
原文高亮摘录
“regularization is a process that converts the answer to a problem to a simpler one.”
Whisper 随想笔记
So basically it's like making a messy solution look neat, even if it's not exact.
分享本页 Whisper
短链接
https://domwhisper.com/s/6d7eeca1bc18嵌入代码
<iframe src="https://domwhisper.com/embed/6d7eeca1bc18" width="100%" height="480" style="border:0;border-radius:16px" loading="lazy"></iframe>