Unverified50% confidenceFactExact time

高偏差意味着模型过于简单导致欠拟合，高方差意味着模型学入噪声导致过拟合，两者往往此消彼长

1

Sources

50%

Confidence

Long-term

Relevance

7/12/2026

First Seen

Sources

机器学习核心概念深度解析：那些让人豁然开朗的顿悟时刻

redditr/learnmachinelearning7/11/2026

Related Claims

Unverified膨胀、过时、充满噪声的评估集比没有评估集更糟糕，因为它会给出误导性的信号79% similar Verified模型的泛化误差等于偏差平方加方差加不可约噪声78% similar Unverified过度膨胀的上下文会稀释模型注意力、抬高推理成本，甚至引入噪声导致幻觉78% similar Unverified扩散模型的迭代去噪过程对数值误差敏感，每步的微小误差会累积并影响最终图像，采样步数越多量化误差影响越大76% similar Verified上下文中的噪声信息会以非线性方式降低模型推理准确性76% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/494147

API

curl https://kongchang.com/api/v1/knowledge/claims/494147

MCP

get_claim(id=494147)