Data augmentation - Wikipedia
6
公开标注数
6
参与人数
2026-07-22 09:41:20
首次 Whisper
本页的公开 Whisper
划选高亮2026-07-22 16:05:20
原文高亮摘录
“Geometric transformations alter the spatial properties of images to simulate different perspectives”
Whisper 随想笔记
Makes sense, but doesn't this just mean rotating and flipping pictures?
划选高亮2026-07-22 13:02:20
原文高亮摘录
“Synthetic Minority Over-sampling Technique (SMOTE) is a method used to address imbalanced datasets”
Whisper 随想笔记
Doesn't this just make the model overfit to fake data though? Seems risky.
划选高亮2026-07-22 12:53:20
原文高亮摘录
“Synthetic Minority Over-sampling Technique (SMOTE) is a method used to address imbalanced datasets”
Whisper 随想笔记
SMOTE saved my butt on a fraud detection project, the minority class was basically invisible.
划选高亮2026-07-22 09:59:20
原文高亮摘录
“Data augmentation is a statistical technique which allows maximum likelihood estimation from incomplete data.”
Whisper 随想笔记
I've used this in Bayesian stuff, it does help with missing values, neat.
划选高亮2026-07-22 09:50:20
原文高亮摘录
“Data augmentation is a statistical technique which allows maximum likelihood estimation from incomplete data.”
Whisper 随想笔记
Wait, isn't this just about making fake data? Seems more like a trick than stats.
划选高亮2026-07-22 09:41:20
原文高亮摘录
“Data augmentation is a statistical technique which allows maximum likelihood estimation from incomplete data.”
Whisper 随想笔记
So that's why we add noise to images — it's basically filling gaps for the model.
分享本页 Whisper
短链接
https://domwhisper.com/s/5e2dc1219aa9嵌入代码
<iframe src="https://domwhisper.com/embed/5e2dc1219aa9" width="100%" height="480" style="border:0;border-radius:16px" loading="lazy"></iframe>