Sarah Lutteropp

Wait... do I really need the fully computed displayed tree loglikelihoods here? For LikelihoodModel.BEST, we can do without explicitly computing them. Because there we essentially have one tree per partition......

Turns out that for networks, we can't go down to a per-site computation basis if we have more than one displayed tree. - I am convinced about this in the...

Okay, found a counterexample (for x > 0): - f(x) = 2x + 3 - g(x) = 2x + 5 - f'(x) = 2 - g'(x) = 2 - max{f'(x),...

This makes it clear that for networks, we need to compute all three of logL(T|P), (logL(T|P))', (logL(T|P))'' with help of the sumtable. The only question remains whether it makes sense...

Computation of second network loglikelihood derivative caused numerical problems (underflow/overflow). Thus, I had to play a bit with the formula: 

Damn the Newton Raphson in pll-modules is awful, also the second network loglikelihood derivative is such a high number that it does not fit into a double. Luckily, what Newton-Raphson...

Or no wait, maybe I can adapt the pll_modules implementation to directly take the quotient as argument...

> Or no wait, maybe I can adapt the pll_modules implementation to directly take the quotient as argument... I tried this (passing the quotient and trimming too large and too...

Mhm. It also happens with my own NR implementation... maybe the derivatives are still wrong?

I checked how Newton-Raphson decides on the next branch length to propose: It essentially subtracts first_logl_derivative/second_logl_derivative from the previous branch length. With these values I get for network first and...