Unverified50% confidenceFactExact time
The standard cross-entropy loss function naturally favors high-frequency patterns during optimization, a phenomenon known in NLP as frequency bias.
1
Sources
50%
Confidence
Long-term
Relevance
9/3/2026
First Seen
Sources
Related Entities
Related Claims
VerifiedCross-entropy loss originates from KL divergence in information theory and measures the degree of difference between the true probability distribution and the predicted probability distribution.70% similarUnverifiedThe probability distribution learned by a model is naturally biased toward high-frequency events and common patterns, resulting in a systematic underestimation of long-tail events.65% similarUnverified扩散模型通过逐步去噪的生成机制产生的频域异常比GAN更微弱且分布不规律，使传统基于频域分析或GAN指纹的检测方法效果大打折扣64% similarUnverified标准交叉熵损失在目标占图像面积较小时会导致严重的类别不平衡问题，模型倾向于全部预测为背景64% similarUnverified如果交叉熵稳步下降而Dice波动剧烈，问题几乎可以确定出在Dice损失的配置上63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/851602API
curl https://kongchang.com/api/v1/knowledge/claims/851602MCP
get_claim(id=851602)