How statement and proof provenance work
The first chip identifies the source of the statement or construction; the second identifies the source of its local proof or verification.
- Literature-sourced: the exact statement appears in a cited source; only wording and notation differ.
- AI-adapted: a semantically identical restatement of literature-sourced material, modulo indexing, notation, and boundary cases adopted by the library.
- AI-generated: a genuinely novel statement formulated by AI, with no source for the claim itself.
These labels describe origin, not correctness: citations and verification chips remain separate evidence.
Smoothed margin monotonicity along homogeneous Clarke flow
Statement
Let and let be locally Lipschitz functions that are positively homogeneous of common degree and admit the Clarke chain rule along absolutely continuous arcs. Set . Let be a global Clarke gradient-flow trajectory for , and suppose for some . For , the loss is nonincreasing, the smoothed normalized margin is nondecreasing, , and . Moreover, , so the normalized margin and smoothed normalized margin have the same finite positive limit.
Statement references
Facts & Assumptions
Given: The finite homogeneous margins, their local Lipschitz and Clarke chain-rule properties, the global Clarke flow, and a time with loss below one.
Clarke subdifferential and gradient-flow trajectory: a Clarke gradient-flow trajectory is locally absolutely continuous, and the subdifferential is the closed convex hull of limiting ordinary gradients, is nonempty compact convex, and has support function the Clarke generalized directional derivative.
Euler identity for Clarke subgradients of homogeneous margins: every satisfies .
Clarke chain rule for finite exponential empirical risk: for sample margins in a homogeneous-classifier model, along a Clarke flow, almost everywhere, and almost everywhere with .
For a nonempty subset of with , compactness, closedness and boundedness, pseudocompactness, and attainment of extrema by every continuous real-valued function are equivalent: continuous real-valued functions attain finite extrema on nonempty closed bounded subsets of finite-dimensional Euclidean space.
Proof
We establish the identities of [F4] for the arbitrary margins allowed here. At any fixed parameter , put and . Since , the two functions have exactly the same differentiability points, and there. Taking gradient limits and closed convex hulls in [F1] gives . Set , a nonempty compact convex set. For any direction , the finite-sum limsup inequality and [F1] give . Thus : otherwise, for , let minimize on , using [F5]. Convexity gives for every , whereas , contradicting the directional bound. Consequently the flow inclusion supplies with almost everywhere. The assumed arc chain rules give almost everywhere. All these compositions are locally absolutely continuous, since the margins are locally Lipschitz. No graph realization is needed.
Put , , and . By [F1], is locally absolutely continuous; local Lipschitz continuity of the margins makes these compositions locally absolutely continuous on each compact time interval where . Since and [step 1.1] gives almost everywhere, for all . Thus every . Homogeneity gives , so and .
At almost every , set , , and . By [step 1.1], choose with . Euler's identity [F3] yields , hence . Also , so because , , and .
The chain rule in [step 1.1] gives . Differentiating and applying Cauchy–Schwarz, , because . Thus the smoothed margin is nondecreasing.
The function is continuous on the unit sphere, so [F5] gives a finite maximum . Homogeneity and imply . Since is nondecreasing by step 4.1, it has a finite positive limit. Writing gives and .
The nonincreasing loss has a limit . If , then and almost everywhere by step 3.1, so while remains bounded. This forces , contradicting the positive lower bound in step 5.1. Hence . To show , suppose otherwise. Then some and times satisfy . The closed ball of radius is compact by [F5], and each continuous is bounded above there. Since , this gives a common positive lower bound for , contradicting . Thus . The error bound in step 5.1 now tends to zero, so the normalized and smoothed margins converge to the same positive limit.
Depends on
- Clarke subdifferential and gradient-flow trajectory
- Homogeneous classifier margins and exponential training risk
- Euler identity for Clarke subgradients of homogeneous margins
- Clarke chain rule for finite exponential empirical risk
- For a nonempty subset of $\mathbb{R}^n$ with $n\ge1$, compactness, closedness and boundedness, pseudocompactness, and attainment of extrema by every continuous real-valued function are equivalent
Used by
Dependency tree · two levels
13 results within two dependency steps of this one, each drawn at its shortest distance from it. An arrow runs from a result to what uses it, so the chart reads left to right and ends at this result, which carries a heavier outline. Every node is a link to that result. Click elsewhere on the chart to enlarge it.
Sources
- Kaifeng Lyu and Jian Li, Gradient Descent Maximizes the Margin of Homogeneous Neural Networks, Theorem 4.1 and Appendix A.7 (standard reference, not scraped)
- Matus Telgarsky, Deep Learning Theory lecture notes, §10.2 (standard reference, not scraped)