All Writings

Expected Calibration Error, Platt scaling, temperature annealing, exchangeability, non-conformity scores, and finite-sample coverage guarantees.

Harmonic functions on graphs, transductive margin bounds, co-training multi-view consensus, and Fisher information in query synthesis.

Spectral graph theory, preserving geodesic distances, stochastic neighbor embedding, and Riemannian geometry in non-linear dimensionality reduction.

Distributions over function spaces, kernelized covariance operators, exact Bayesian posterior conditioning, and active Bayesian optimization.

Converting intractable integration into deterministic optimization: Mean-field factorizations, Kullback-Leibler geometry, and coordinate ascent variational inference (CAVI).

Maximum likelihood in the presence of unobserved variables: The Expectation-Maximization algorithm, variational lower bound dynamics, and Gaussian Mixture Models.

Conditional independence on topological manifolds: Directed Bayesian networks, undirected Markov random fields, the Hammersley-Clifford theorem, and Judea Pearl’s d-separation calculus.

From PAC weak learnability to gradient descent in infinite-dimensional function manifolds: Robert Schapire’s AdaBoost, Jerome Friedman’s gradient machines, and Tianqi Chen’s XGBoost.

Taming instability through parallel stochastic aggregation: Bradley Efron’s bootstrap, Leo Breiman’s bagging, pairwise tree decorrelation, and the asymptotic mechanics of Random Forests.

Non-metric hierarchical classification: Claude Shannon’s entropy, Gini concentration impurity, greedy recursive partitioning, and cost-complexity pruning in decision trees.