Good to know

Conditional Multivariate Normals

using Distributions
using LinearAlgebra

##

# P(A|B)

Σab = 0.2*randn(3,3)
Σab += Σab'
Σab += diagm([1.0;1.0;1.0])

μ_ab = [10.0;0.0;-1.0]
μ_1 = μ_ab[1:1]
μ_2 = μ_ab[2:3]

Σ_11 = Σab[1:1,1:1]
Σ_12 = Σab[1:1,2:3]
Σ_21 = Σab[2:3,1:1]
Σ_22 = Σab[2:3,2:3]

##

# P(A|B) = P(A,B) / P(B)
P_AB = MvNormal(μ_ab, Σab)                       # likelihood
P_B = MvNormal([-0.5;0.75], [0.75 0.3; 0.3 2.0]) # evidence

# Schur compliment
μ_(b) = μ_1 + Σ_12*Σ_22^(-1)*(b-μ_2)
Σ_ = Σ_11 + Σ_12*Σ_22^(-1)*Σ_21

P_AB_B(a,b) = pdf(P_AB, [a;b]) / pdf(P_B, b)
P_A_B(a,b; mv = MvNormal(μ_(b), Σ_)) = pdf(mv, a) 

##

# probability density: p(a) = P(A=a | B=b)
@show P_A_B([1.;],[0.;0.])
@show P_AB_B([1.;],[0.;0.])

P(A|B=B(.))

Various Internal Function Docs

IncrementalInference._solveCCWNumeric!Function
_solveCCWNumeric!(ccwl; ...)
_solveCCWNumeric!(ccwl, _slack; perturb)

Solve free variable x by root finding residual function fgr.usrfnc(res, x). This is the penultimate step before calling numerical operations to move actual estimates, which is done by an internally created lambda function.

Notes

  • Assumes cpt_.p is already set to desired X decision variable dimensions and size.
  • Assumes only ccw.particleidx will be solved for
  • small random (off-manifold) perturbation used to prevent trivial solver cases, div by 0 etc.
    • perturb is necessary for NLsolve (obsolete) cases, and smaller than 1e-10 will result in test failure
  • Also incorporates the active hypo lookup

DevNotes

  • TODO testshuffle is now obsolete, should be removed
  • TODO perhaps consolidate perturbation with inflation or nullhypo
source