Unverified50% confidenceFactExact time
反向传播机制要求整个计算图必须是可微分的，且在极深网络中存在梯度消失/爆炸问题
1
Sources
50%
Confidence
Long-term
Relevance
7/21/2026
First Seen
Sources
Related Claims
Unverified数值微分计算量与参数数量成正比，符号微分会产生表达式膨胀问题，二者在大规模网络中都无法扩展，反向传播是解决方案74% similarUnverifiedVanishing gradient refers to the exponential decay of gradient values during backpropagation in deep neural networks, causing parameters near the input layer to receive almost no updates.73% similarUnverified级联架构面临的核心挑战是误差传播，上游模型的偏差会被放大传递到下游70% similarUnverified状态离散化存在权衡：分桶粒度过粗导致信息损失，过细则引发维度灾难，这也是DQN用神经网络替代Q表的核心动机69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/577011API
curl https://kongchang.com/api/v1/knowledge/claims/577011MCP
get_claim(id=577011)