Unverified50% confidenceFactExact time
前景理论的核心发现是损失厌恶，损失曲线的斜率远陡于同等幅度收益曲线的斜率
1
Sources
50%
Confidence
Long-term
Relevance
7/21/2026
First Seen
Sources
Related Claims
VerifiedGradient descent iteratively adjusts weights in the opposite direction of the loss function's gradient, causing the prediction error to converge progressively.79% similarUnverified梯度下降的收敛条件与学习率和损失函数的Lipschitz常数的关系有关72% similarUnverified香农的率失真理论证明，在给定比特率约束下，任何有损压缩都存在不可逾越的失真下界70% similarUnverifiedSanturkar等人2018年的研究指出BN的真正作用可能在于平滑损失曲面而非减少内部协变量偏移69% similarUnverifiedThe core loss functions of contrastive learning (such as InfoNCE and NT-Xent) essentially pull the representations of similar samples closer together while pushing the representations of dissimilar samples further apart.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/577545API
curl https://kongchang.com/api/v1/knowledge/claims/577545MCP
get_claim(id=577545)