REVIEW 6 cited by
A Slightly Improved Bound for the KLS Constant
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
read the original abstract
We refine the recent breakthrough technique of Klartag and Lehec to obtain an improved polylogarithmic bound for the KLS constant.
Forward citations
Cited by 6 Pith papers
-
Affirmative Resolution of Bourgain's Slicing Problem using Guan's Bound
Bourgain's slicing problem is resolved: every volume-one convex body in R^n has a hyperplane section of (n-1)-volume at least a universal constant c > 0.
-
The KLS constant is $O(\log^{1/4} n)$
Every isotropic log-concave measure satisfies a quadratic-form Poincaré inequality with constant 2, which implies the KLS constant is at most C log^{1/4} n.
-
Digesting the proof of the sharp thin-shell inequality
The optimal constant in the log-concave variance conjecture is E[(|X|²−n)²] ≤ 8n, attained by iid standard centered exponentials; a companion sharp bound holds for third moments and for the simplex among convex bodies.
-
Regularized Dikin Walks for Sampling Truncated Logconcave Measures, Mixed Isoperimetry and Beyond Worst-Case Analysis
The soft-threshold Dikin walk mixes in O((m+kappa)n) iterations for truncated logconcave targets, supported by a new isoperimetric inequality combining Euclidean and Hilbert metrics.
-
Improved generalization bounds for binary linear classification via isoperimetry
Uniform generalization errors in binary linear classification concentrate around their expectation at O(1/sqrt(n)) rates under unbounded Lipschitz losses, via new log-Sobolev inequalities for (Y_i X_i, Y_i).
-
On the Statistical Capacity of Deep Generative Models
Push-forwards of Gaussian or log-concave latent variables through Lipschitz neural networks are always sub-Gaussian or sub-exponential, so common deep generative models cannot generate heavy-tailed distributions.
Discussion (0). Continue with ORCID to comment.