HybridGaussianConditional inherits from HybridGaussianFactor

[varunagrawal]

Analogous to GaussianConditional and GaussianFactor, make HybridGaussianConditional inherit from HybridGaussianFactor so we can reduce code duplication.

I also changed logConstant_ to be defined in the negative logspace, letting us lower bound the values at 0.0 instead of $-\infty$. I'll be updating the Hybrid.pdf file to reflect this for synergy.

One of a bunch of small PRs to improve the hybrid API and functionality.

[dellaert]

LG, but this makes me think that maybe all log constants should be in negative log space. Either k, or -log k. Never log k.

[varunagrawal]

LG, but this makes me think that maybe all log constants should be in negative log space. Either k, or -log k. Never log k.

Yeah I have been thinking about that as well.

[dellaert]

LG, but this makes me think that maybe all log constants should be in negative log space. Either k, or -log k. Never log k.

Yeah I have been thinking about that as well.

I think we should do it. Will you or will I?

[varunagrawal]

I can get to it later tonight, if that works?

[varunagrawal]

Just to make sure I understand, the change is to update logNormalizationConstant so that it returns -log(k) where $k = \frac{1}{\sqrt{|2\pi \Sigma|}}$ right?
Currently that is the only method/function that works with log(k).

[dellaert]

Yes. But there is one (or more?) in HC class hierarchy and one in noiseModel