Unverified50% confidenceFactExact time
The proximal operator prox_{ηh}(v) = argmin_x { h(x) + (1/(2η))||x-v||^2 } is equivalent to performing one step of implicit gradient descent on h at the point v, and does not require h to be differentiable.
1
Sources
50%
Confidence
Long-term
Relevance
10/12/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe general form of bilevel optimization is min_θ F(θ, φ*(θ)), where φ*(θ) = argmin_φ G(θ, φ), and its gradient can be computed efficiently via the implicit function theorem or iterative differentiation techniques.67% similarUnverifiedThe typical form of Zeller's formula is h = (q + ⌊13(m+1)/5⌋ + K + ⌊K/4⌋ + ⌊J/4⌋ - 2J) mod 7.65% similarUnverifiedThe residual block formula is y = x + F(x); during backpropagation, the gradient passes through the "1" in (1 + dF/dx), providing a shortcut for gradients to flow directly to earlier layers.64% similarUnverified一维Wasserstein-1距离等于两个累积分布函数之间面积的积分 W₁ = ∫|F_A(x) - F_B(x)|dx63% similarUnverifiedThe mathematical operation for ablating the refusal direction is a' = a - (a · r̂) r̂, which orthogonally projects the activation vector onto a hyperplane orthogonal to the refusal direction.62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1012938API
curl https://kongchang.com/api/v1/knowledge/claims/1012938MCP
get_claim(id=1012938)