Closed RodenLuo closed 3 months ago
Fit search is supposed to be reproducible. As with any scientific computation, even ones involving random choices, it is important to be able to exactly reproduce results. So I think the behavior you see is correct. You may want to have an option to set the random seed to get different results.
Best to report issues such as this using ChimeraX menu Help / Report a Bug so it goes to our Trac database.
(Encountered "Error while submitting feedback." Reported via Email.)
The other non-random case is with PDB-8GAM and EMD-29900.
If one tries the same thing with PDB-8TNP and EMD-41423, then the inits seem to be random and not reproducible.
Important parts of the log are attached at the end. I have another random example, but the input data is a bit hard to describe (can send them to you if needed). My general feeling is when the target volume's box size is small, the inits are random. When the box is large, the inits are reproducible.
I expected the default to be fully random at all times unless the user specifies a random seed. Even with a random seed, I expected the inits to be different with different N (searching times). But this is more of a personal preference. To be consistent should be by design I guess.
I will investigate in April. The behavior should not vary based on the data and reproducible or random should be bith possible based on a seed parameter.
I've made a ticket in our ChimeraX Trac database for this since that is where ChimeraX bug reports are handled.
https://www.rbvi.ucsf.edu/trac/ChimeraX/ticket/14863
Sorry you got "Error while submitting feedback" when trying to report the problem in ChimeraX. That happens when our UCSF web server that receives the reports is down.
Hi,
The following lines returned deterministic values in one case for my settings. I suppose they should be random and this should not happen by design.
https://github.com/RBVI/ChimeraX/blob/4d2ef439300bfa2587e272a0b0de85382e2b5993/src/bundles/map_fit/src/search.py#L71-L74
To reproduce
On Windows 11, ChimeraX version 1.7.1 (2024-01-23)
fit_search
def as fit_search.txt. No change in the logic, only added debugging log output.PDB-8JGF
andEMD-36232
runscript "PATH to fit_in_chimerax.py" "PATH to 8jgf.cif" "PATH to any valid folder for a log output, such as the Downloads folder" "PATH to emd_36232.map" 0.20 2.7 3
Out of 3 runs, I got the following log. The random variables are the same. This is reproducible on different days (I first thought there might be a seed using the date). This is not happening for any other cases I have performed so far.
I tried to open this at Trac (totally new to me), but seemed not possible for me to do so. So, I'm still posting here...
Thanks, Roden