Generalization error bounds for two-layer neural networks with Lipschitz loss function

Collection of recent theoretical and methodological ML papers (arXiv 2604.*) covering: generalization bounds for two-layer networks using Wasserstein estimates; robust distributed mean estimation with adversarial/async workers and tight finite-time rates; variational clustering for noisy high-dimensional data; amortized filtering/smoothing with conditional normalizing flows; scalable nearest-neighbour Gaussian process theory; high-dimensional CLT rates for asynchronous Q-learning; SGD dynamics in saddle-to-saddle regimes; robustness analysis and code for spiking-reservoirs; Weighted Bayesian (