Unverified50% confidenceFactExact time
Even when approximating only a single fixed affine function, at a sufficiently high confidence level, the cost of the output layer must reach ((1-p)/p)ε^{-2}log(1/δ).
1
Sources
50%
Confidence
Long-term
Relevance
10/5/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe lower-bound argument comes from two perspectives: Sobolev capacity constraints and the output-layer cost for approximating a fixed affine function.72% similarUnverifiedFor fixed p∈(0,1) and δ<min{1/2,1-p}, under a fixed or logarithmic depth budget, the upper bound and lower bound match exactly in the accuracy exponent.69% similarUnverifiedScore函数定义为对数密度梯度 ∇x log p(x)，指向概率密度增大最快的方向，其对数梯度使归一化常数消失62% similarUnverifiedδ*(r) is the inverse rate-distortion function of a uniform K-ary source under zero-one loss, characterizing the distortion that inevitably arises when each fact has only B/M bits of storage.60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/975786API
curl https://kongchang.com/api/v1/knowledge/claims/975786MCP
get_claim(id=975786)