High-Dimensional Statistics

Read these first

1. 01Concentration of Measure in High Dimensions

   In high dimensions, quantities you'd expect to spread out widely instead pile up tightly around their average, the same fact that makes big averages reliable also makes high-dimensional geometry deeply unintuitive.

2. 02The p Greater Than n Regime

   When you have more candidate features than observations, ordinary regression doesn't just get noisier, it stops having a unique answer at all, and that changes everything about how you have to estimate.

Then the rest

1. 03Knockoff Filters for Controlled Variable Selection
2. 04Nonlinear Shrinkage of Covariance Eigenvalues
3. 05Randomized SVD and Sketching for Large Matrices
4. 06Sparsity and the Bet-on-Sparsity Principle
5. 07Stability Selection
6. 08Sub-Gaussian and Sub-Exponential Tail Bounds
7. 09Sure Independence Screening
8. 10Spiked Covariance Models and the BBP Transition