Sarah Lutteropp

That sounds even better. I've taken a look at the ancestral state computation for pll_utree_t, adapting this to the network displayed trees looks pretty straight-forward. If we find a way...

I don't see yet how the ancestral states for the displayed trees of a r-reticulation network help in finding the arc insertion move to take in order to get to...

Found the old notes: https://github.com/lutteropp/NetRAX/issues/30#issuecomment-741810684 The idea is to find two nodes whose ancestral states have the smallest distance, but are located in different places in the network. Makes sense...

Sounds like we can simply reuse the fake-node-approach here. (adding the fake node as the second child with brlen zero, the fake node having all-equal ancestral state probabilities) The fake-node-approach...

This should work: - for each network node u, we store a set AS(u) of ancestral states (one per displayed tree) - the ancestral distance anc_dist(u, v) between two network...

A possible definition for dist((u,s1),(v,s2)) would be: (1.0 - dist(s1,s2)) * dist(u,v). Here, I assume that dist(s1, s2) is in range [0,1]. If it is not, we can normalize it...

It is still unclear to me whether dist(u,v) should be the *number of edges* on a edge-minimal path between u and v in the network, or whether the *sum of...

Great, it works! :+1: It is very slow, but it works! (The slightly adapted version is in https://github.com/lutteropp/NetRAX/blob/master/simulator/src/network/logic/celine_simulator.py) Due to runtime issues, it is not realistic to exactly pre-specify wanted...

There is one problem with our simulation setup from above - it ignores the displayed tree probabilities. I implemented the following sampling types: ``` class SamplingType(Enum): STANDARD = 1 #...

Okay, switched to PERFECT_SAMPLING as we all agree on that. Done in https://github.com/lutteropp/NetRAX/commit/addb906a4041f26dbba2e50be87fc2546e13d7b3. (EDIT: Moved starting networks question to [new issue](https://github.com/lutteropp/NetRAX/issues/7))