MW Math Workbench

deterministic kernels · LP · Kalman · Bayes · SDE · AR · GARCH (phases 275-279) · all bit-exact replay

◂ MC Mission Control

LP Linear Program (Phase 275)

min/max c·x s.t. A·x {=,≤,≥} b · revised simplex + Big-M

sense

objective coeffs (csv)

constraints (one per line: "coeffs|kind|rhs", kind=<=,>=,=)

— ready —

— press ▸ Solve —

KF Kalman filter (Phase 276)

linear-Gaussian state-space · predict/update over observations

n state dims

m obs dims

F (row-major csv)

H (row-major csv)

Q diag (csv)

R diag (csv)

x0 (csv)

P0 diag (csv)

observations (one per line, csv of m values)

— ready —

— press ▸ Run filter —

BN Bayesian network (Phase 277)

exact inference · P(target | evidence) · CPT-validated

network JSON (variables + CPTs)

target variable evidence (k=v csv)

— Russell&Norvig burglar/earthquake — expect P(B=t)≈0.284 —

— press ▸ Infer —

SD SDE Monte Carlo (Phase 278)

dX = μ(X)·dt + σ(X)·dW · Euler-Maruyama / Milstein · seeded PCG

model

X₀

μ (drift / OU mean)

θ (OU only)

T₀ → T₁

steps

npaths

seed

scheme

— ready —

— press ▸ Run MC —

AR AR(p) fit (Phase 279)

Yule-Walker + Levinson-Durbin · recovers φ, σ, mean

order p

series (csv of floats — paste, or use sample button)

— ready —

— press ▸ Fit —

GH GARCH(1,1) fit (Phase 279)

σ²ₜ = ω + α·r²ₜ₋₁ + β·σ²ₜ₋₁ · coordinate-descent MLE · α+β<1

returns (csv of floats — paste, or sample)

— ready —

— press ▸ Fit —