search

aseyboldt / covadapt-paper

1 stars 1 forks source link

LaTeX and PDF out of sync #1

Open maresb opened 17 hours ago

maresb commented 17 hours ago

Seems to me like the PDF is best thought of as an artifact rather than being checked in. I've been meaning to look into GHA workflows, but haven't gotten around to it.

maresb commented 17 hours ago

I'm writing because I really liked "Motivation: Example with an independent gaussian posterior" but didn't find it in the compiled version.

I wanted to correct this:

image

I was sure that the formula for $\sigma^2$ had to be wrong since it looks so weird with a square on the LHS and square root on the RHS, but shockingly-to-me it's correct.

maresb commented 16 hours ago

Oh, it's obvious from dimensional analysis since $\alpha_i$ has units of $1/x$ and $\sigma$ has units of $x$.

aseyboldt commented 15 hours ago

I switched to typst, so the source is in main.typ. I guess I should delete a lot of unused files...

aseyboldt commented 15 hours ago

And maybe it is more obvious why the formula would look like this if you think about it as $\sigma^2 = \text{geomean}(Var[x_i], Var[\alpha_i]^{-1}) = \exp(\log Var[x_i] / 2 - \log Var[\alpha_i] / 2)$.