Lipschitz Continuity in Deep Learning: A Systematic Review of Theoretical Foundations, Estimation Methods, Regularization Approaches, and Certifiable Robustness

A batch of new arXiv stat-ml submissions covering theoretical and methodological advances across deep learning and statistical learning. Key items: a comprehensive survey on Lipschitz continuity in neural networks (theory, estimation, regularization, certifiable robustness); Meta-Thresholding for semi-supervised learning that learns pseudo-label thresholds; Split Forward Gradients for backpropagation-free trunk training with variance-reduction and memory savings; Isotonic Conformal Prediction (split and transductive variants) for self-calibration and prediction-conditional validity; label-aug-