somewhere near quantisation: mixture of experts (MoE) compression page on Mixture of Experts models generally? + confirm - are these specific to LLMs, or neural networks more generally? see to be more general?
LightGBM Light Gradient Boosting Machine
gram matrix to get kernel (stats stuff) page on boosting. xgboost? page on one-shot, zero-shot and few-shot learning.
ordinary linear regression for inference split out: + gauss markov
Point variable estimates for discriminative models: + split out
page on Multilevel regression with poststratification (MRP)
split out ensemble methods: + boosting and bagging
something on additive models around non-parametric regression
split out non-parametric regression. many of those should be in linear
xlearner tlearner
ordinary least squares for prediction split out + frisch-waugh-lovell + bayesian linear regression to bayrsian bit above + least trimmed squares + linear models + impact of outliers (leverage and cook’s distance)
page on feature engineering