Unverified50% confidenceFactExact time
When the step size $\eta \to 0$, the trajectory of the SGD iteration $x_{k+1} = x_k - \eta \nabla f(x_k) + \eta \xi_k$ can be approximated by an SDE: $dX_t = -\nabla f(X_t)\,dt + \sqrt{\eta}\,\Sigma\,dW_t$.
1
Sources
50%
Confidence
Long-term
Relevance
10/14/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe general form of a stochastic differential equation is dX = f(X,t)dt + g(X,t)dW, where f is the drift term and g is the diffusion term.68% similarPartially VerifiedThe mathematical form of KL divergence is D_KL(p||q) = ∑ p(x_i) log(p(x_i)/q(x_i)).67% 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.67% similarUnverifiedThe push-forward curve is defined as ρ_t = ((1-t)id + tT)_#ρ, whose tangent vector at t=0 is T-id, corresponding to the displacement direction in Wasserstein space64% similarUnverified墨卡托投影的坐标变换公式为 x = R·λ，y = R·ln[tan(π/4 + φ/2)]63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1026660API
curl https://kongchang.com/api/v1/knowledge/claims/1026660MCP
get_claim(id=1026660)