Marginalization of variables in an incremental update setup

[chartl]

@pekoto

Was it supposed to be this instead:

P^{(\omega)}(l,t_\omega,\theta|D_\omega)\propto P(l) P(t_\omega | \theta) P(\theta) P(D_\omega | l, t_\omega) P(ω)(l,tω,θ|Dω)∝P(l)P(tω|θ)P(θ)P(Dω|l,tω)

Yes

@pekoto

this is how the likelihood roughly looks like:

This doesn’t really make clear the dependence over l. You can write

L(m|u, c_i, l) = L(m|u, c_i, 0)l + L(M|u, c_i, 1)(1-l), but if the L() terms don’t simplify it will be hard to integrate away l.