THE FINTECH BUILDER · SYNTHETIC LEARNING LAB

Ledoit-Wolf Shrinkage

Compare estimated outer-product noise with target distance.

0.7
1

Keyboard: Tab to a control; native arrow keys change sliders. Alt + Right steps, Alt + Left goes back. Reduced-motion Play advances one state only.

  1. 1. Center data and divide by n
  2. 2. Build average-variance identity target
  3. 3. Estimate noise / target distance
  4. 4. Clip intensity and blend matrices

Calculated history

Calculated stateValues from the current input prefix, not a decorative trace.
● Selected experiment┄ Canonical parameters / same cutoff

Result and diagnostics

C=sum(z_i z_i^T)/n; mu=tr(C)/p; beta=sum||z_i z_i^T-C||_F²/n²; rho=clip(beta/||C-mu I||_F²,0,1)

Calculation trace

Exact current output (JSON)


Input audit
Only the shown information prefix is supplied to the calculation. Synthetic fixture; no live market data or fitted performance claim.
Exact input supplied to the reference engine