Unverified50% confidenceFactExact time
The state-transition Jacobian matrix characterizes the sensitivity of the RNN hidden state to the hidden state at the previous time step, and the norm behavior of its product determines whether gradients can be effectively propagated to distant points.
1
Sources
50%
Confidence
Long-term
Relevance
11/6/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedRNN通过隐藏状态在时间步上串行传递信息，导致梯度消失/爆炸和无法并行计算两个核心瓶颈74% similarUnverifiedRNN 引入隐状态机制，理论上能记忆任意长度上下文，但受梯度消失问题困扰72% similarUnverifiedThe Jacobian lens compresses the matrix of partial derivatives of an intermediate-layer hidden state with respect to the final output logits into low-dimensional coordinates indexed by vocabulary, decoupling the communication signal from model width.72% similarUnverifiedUnder the assumption that the state transition Jacobian decays exponentially, the error between the truncated gradient and the full-context gradient decays exponentially as the distance from the context boundary increases.70% similarUnverifiedNRALD targets sampling from non-differentiable distributions on constrained spaces, employing a normal reflection mechanism at the boundary.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1141543API
curl https://kongchang.com/api/v1/knowledge/claims/1141543MCP
get_claim(id=1141543)