Proves GD convergence to stationary point neighborhoods for general NN architectures beyond NTK via block-level analysis, analyticity, and local smoothness conditions.
The Zero Set of a Real Analytic Function
10 Pith papers cite this work. Polarity classification is still indexing.
10
Pith papers citing it
abstract
A brief proof of the statement that the zero-set of a nontrivial real-analytic function in $d$-dimensional space has zero measure is provided.
citation-role summary
background 1
citation-polarity summary
roles
background 1polarities
background 1representative citing papers
Causal ATE/CATE are identifiable for categorical unobserved confounders from three or more conditionally independent proxies or treatments, recovered consistently by tensor decomposition of the mixture.
A diameter criterion tied to a potential function certifies convergence of difference inclusions, enabling discrete proofs for first-order optimization methods with diminishing steps.
The authors propose an S-MILP framework that optimizes group sequential testing boundaries to achieve faster rejection of the null hypothesis compared to traditional methods while controlling type I and type II errors.
Ill-chosen reference states induce spurious long-range strange correlators in trivial SPT phases through magnitude degeneracy of the transfer matrix in MPS representations.
Derives information-theoretic recovery thresholds for two intersecting lines with polynomial mass concentration near the intersection and matches them (up to polylog factors) via a spectral algorithm on a hypergraph built from nearly collinear triples.
Identifiability is proven for recurrent nonlinear switching dynamical systems under flexible assumptions, and ΩSDS is introduced as a flow-based estimator that improves disentanglement and forecasting over VAE-based methods.
Polynomial representations yield an effective-degree simplicity metric that predicts generalization across tasks and serves as a differentiable regularizer improving performance in classification and RL.
MMD GANs have unbiased critic gradients but biased generator gradients from sample-based learning, and the Kernel Inception Distance provides a practical new measure for GAN convergence and dynamic learning rate adaptation.
Explicit constructions show that isospectral graphons can arise from distinct geometries and are not combinatorially equivalent, with mixed implications for stability in graphon Kuramoto dynamics.
citing papers explorer
-
Convergence of Gradient Descent for General Neural Network Architectures Beyond the NTK Regime
Proves GD convergence to stationary point neighborhoods for general NN architectures beyond NTK via block-level analysis, analyticity, and local smoothness conditions.
-
Causal Inference with Categorical Unobserved Confounder via Mixture Learning
Causal ATE/CATE are identifiable for categorical unobserved confounders from three or more conditionally independent proxies or treatments, recovered consistently by tensor decomposition of the mixture.
-
Convergence of difference inclusions via a diameter criterion
A diameter criterion tied to a potential function certifies convergence of difference inclusions, enabling discrete proofs for first-order optimization methods with diminishing steps.
-
A General Framework for Optimal Group Sequential Testing via Mixed-Integer Linear Programming
The authors propose an S-MILP framework that optimizes group sequential testing boundaries to achieve faster rejection of the null hypothesis compared to traditional methods while controlling type I and type II errors.
-
Spurious Strange Correlators in Symmetry-Protected Topological Phases
Ill-chosen reference states induce spurious long-range strange correlators in trivial SPT phases through magnitude degeneracy of the transfer matrix in MPS representations.
-
Consistent line clustering using geometric hypergraphs
Derives information-theoretic recovery thresholds for two intersecting lines with polynomial mass concentration near the intersection and matches them (up to polylog factors) via a spectral algorithm on a hypergraph built from nearly collinear triples.
-
End-to-End Identifiable and Consistent Recurrent Switching Dynamical Systems
Identifiability is proven for recurrent nonlinear switching dynamical systems under flexible assumptions, and ΩSDS is introduced as a flow-based estimator that improves disentanglement and forecasting over VAE-based methods.
-
Quantifying and Optimizing Simplicity via Polynomial Representations
Polynomial representations yield an effective-degree simplicity metric that predicts generalization across tasks and serves as a differentiable regularizer improving performance in classification and RL.
-
Demystifying MMD GANs
MMD GANs have unbiased critic gradients but biased generator gradients from sample-based learning, and the Kernel Inception Distance provides a practical new measure for GAN convergence and dynamic learning rate adaptation.
-
Graphons, Geometry, and Dynamics: Forward and Inverse Perspectives
Explicit constructions show that isospectral graphons can arise from distinct geometries and are not combinatorially equivalent, with mixed implications for stability in graphon Kuramoto dynamics.