DKL(P[X]||∏i≤kP[Xi−1|Xi])P[Xk](∏k≤iP[Xi+1|Xi])=DKL(P[X]||∏i≤k+1P[Xi−1|Xi])P[Xk+1](∏k+1≤iP[Xi+1|Xi])
DKL(P[X]||∏i≤kP[Xi−1|Xi])P[Xk](∏k≤iP[Xi+1|Xi])
=DKL(P[X]||∏i≤k+1P[Xi−1|Xi])P[Xk+1](∏k+1≤iP[Xi+1|Xi])
I think the parentheses are off here. IIUC you want to express the equality of divergences, not divergences multiplied by probabilities (which wouldn’t make sense I think).
ϵblanket geqDKL(P[X,Y,Z]||P[X|Y]P[Y]P[Z|Y])=DKL(P[X,Y,Z]||P[X,Y]P[Z,Y]/P[Y])
ϵblanket geqDKL(P[X,Y,Z]||P[X|Y]P[Y]P[Z|Y])
=DKL(P[X,Y,Z]||P[X,Y]P[Z,Y]/P[Y])
Typo: geq → ≥
Man, that top one was a mess. Fixed now, thank you!
I think the parentheses are off here. IIUC you want to express the equality of divergences, not divergences multiplied by probabilities (which wouldn’t make sense I think).
Typo: geq → ≥
Man, that top one was a mess. Fixed now, thank you!