Proposal: user-friendly fix for setting the random seed for registration

[stnava]

@yarikoptic made a good suggestion that we should use an environment variable to set the random seed for registration in ANTs. e.g.

export ANTS_GLOBAL_DEFAULT_REGISTRATION_SEED=1

this would obviate user-interface code and would allow near immediate porting to ANTsR and ANTsPy.

@cookpa --- any chance you could to take a stab at implementation via pull request? i imagine that there are relatively few places (antsAI, antsMotionCorr, antsRegistration that would need a (pseudo-code)

registration->reinitializeSeed( std::getenv( " Name Here " ) )

or whatever the right ITK way is to do such a thing.

[stnava]

@ntustison any thoughts on this?

[ntustison]

Yeah, I think it's a good idea. I'm not sure about the environmental variable, though. We had a similar issue with Atropos and added a command line option. As another possibility to mitigating this randomization issue, could we add a command line parameter to these other programs? Not sure which route would be best but it would probably be worth considering standardizing one approach.

[satra]

my preference would be for a command line option.

[cookpa]

Yes I think a command line option would be best - I can work on this next week.

…

Sent from my iPhone On Jun 1, 2018, at 5:08 PM, Satrajit Ghosh <notifications@github.com<mailto:notifications@github.com>> wrote: my preference would be for a command line option. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub<#596 (comment)>, or mute the thread<https://github.com/notifications/unsubscribe-auth/AAlUs963Qsa50AnO-c0n0bQO309aD0Uvks5t4a0ugaJpZM4UXTpf>.

[stnava]

Ok ... I am impartial to which solution as long as is consistent.

…

On Fri, Jun 1, 2018, 17:11 Philip Cook ***@***.***> wrote: Yes I think a command line option would be best - I can work on this next week. Sent from my iPhone On Jun 1, 2018, at 5:08 PM, Satrajit Ghosh ***@***.*** ***@***.***>> wrote: my preference would be for a command line option. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub< #596 (comment)>, or mute the thread< https://github.com/notifications/unsubscribe-auth/AAlUs963Qsa50AnO-c0n0bQO309aD0Uvks5t4a0ugaJpZM4UXTpf>. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#596 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AATyfn4UwRxmXoOMLGRJSI-LmuxB1vqsks5t4a30gaJpZM4UXTpf> .

[stnava]

#597 allows envt variable control of seed and can be over-ridden by a command line option when it becomes available.

[cookpa]

I made a start with a command line option to antsRegistration:

792d673

FYI SyN does call the PrepareRegistration method, so it would not get a fixed seed in the current master. This only matters if you are not using dense sampling for SyN though.

I will work on other executables / scripts next.

[stnava]

Nice. I am thinking we make a release after this gets done.

…

On Fri, Jun 8, 2018, 17:03 Philip Cook ***@***.***> wrote: I made a start with a command line option to antsRegistration: 792d673 <792d673> FYI SyN does call the PrepareRegistration method, so it would not get a fixed seed in the current master. This only matters if you are not using dense sampling for SyN though. I will work on other executables / scripts next. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#596 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AATyfkXFwW2vCTY_lQbhbfP9r6WopUfvks5t6ua4gaJpZM4UXTpf> .

[cookpa]

#599