Derives joint asymptotic jump-diffusion limit for global parameters and latent variables in SGLD-Gibbs under space-time rescaling, yielding explicit hyperparameter tuning guidance for calibrated uncertainty quantification.
Hitting the high-dimensional notes: An ode for sgd learning dynamics on glms and multi-index models, 2023
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
representative citing papers
At critical learning rate 1/N, the rescaled correlation of SGD in single-index models is claimed to converge to a mean-reverting Ornstein-Uhlenbeck process, but the diffusion coefficients are internally inconsistent.
citing papers explorer
-
Large-scale Uncertainty Quantification for Latent Variable Models Using Subsampling Markov Chain Monte Carlo
Derives joint asymptotic jump-diffusion limit for global parameters and latent variables in SGLD-Gibbs under space-time rescaling, yielding explicit hyperparameter tuning guidance for calibrated uncertainty quantification.
-
Limit Theorems for Stochastic Gradient Descent in High-Dimensional Single-Layer Networks
At critical learning rate 1/N, the rescaled correlation of SGD in single-index models is claimed to converge to a mean-reverting Ornstein-Uhlenbeck process, but the diffusion coefficients are internally inconsistent.