Parameter presets for improved user-friendliness

[jacione]

Having some "preset" keys in the configuration files might make cohere more approachable for non-experts. For example, instead of

shrink_wrap_trigger = [1, 1]
shrink_wrap_type = "GAUSS"
shrink_wrap_threshold = 0.15
shrink_wrap_gauss_sigma = 1.0

you could just put

shrink_wrap_preset = "standard"

The config loader would see that, pull the relevant parameters from a dictionary, and save them into the configuration object and conf_rec (or whichever file they were in). That way, you could still open conf_rec and see all the actual parameters.

For the GUI, this would probably be best as a drop-down menu. When a preset is selected, it would trigger the relevant input boxes to update with the values from that preset.

As usual, I'm happy to tack this onto the list of things I might eventually get to.

[jacione]

I'm particularly thinking of some of the multipeak features, which are getting increasingly parameter-heavy.

[brussel13]

The gui already has a defaults button for most of the features. I think the example experiment should be distributed with config files that include all of the parameters. Just comment out all but the most basic. I think our documentation could benefit from a “getting started” section. Everything is reasonably well documented. But a short section that describes the steps to set up an experiment from scratch would probably be good. Could also create some basic documentation about how to implement a new beamline. That might already exist in the development section now. We should also work on some simple beamline implementations. The current 34-ID-C is very complicated now so it’s not a good example to follow for setting up a new beamline. We have some data that is from simple geometries. Specular vertical scattering reflections and horizontal scattering. So we could make simple examples. We could also create a 34idc simple that just has the geometry and basic data loading without any parallelizing or multi scan features. I’m also working on diffractometer simulations now and can start making simulated data from an arbitrary diffraction geometry. So we could include those in the examples. Get Outlook for iOS<https://aka.ms/o0ukef>

…

________________________________ From: Nick Porter ***@***.***> Sent: Wednesday, November 13, 2024 6:25:39 PM To: AdvancedPhotonSource/cohere ***@***.***> Cc: Subscribed ***@***.***> Subject: Re: [AdvancedPhotonSource/cohere] Parameter presets for improved user-friendliness (Issue #37) I'm particularly thinking of some of the multipeak features, which are getting increasingly parameter-heavy. — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you are subscribed to this thread. Message ZjQcmQRYFpfptBannerStart This Message Is From an External Sender This message came from outside your organization. ZjQcmQRYFpfptBannerEnd I'm particularly thinking of some of the multipeak features, which are getting increasingly parameter-heavy. — Reply to this email directly, view it on GitHub<https://urldefense.us/v3/__https://github.com/AdvancedPhotonSource/cohere/issues/37*issuecomment-2474276787__;Iw!!G_uCfscf7eWS!chMlJWBUGo2LMhatxy5Hh1QfINLfAJT3_IVMoteGQVU2WmNaJPF7Ok6c5l5ygV45204Z6mRSDzcWosyOoz1EIUGfag$>, or unsubscribe<https://urldefense.us/v3/__https://github.com/notifications/unsubscribe-auth/ABZ6SOESTKG5DGMLXJC6IQT2AODRHAVCNFSM6AAAAABRW773RWVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDINZUGI3TMNZYG4__;!!G_uCfscf7eWS!chMlJWBUGo2LMhatxy5Hh1QfINLfAJT3_IVMoteGQVU2WmNaJPF7Ok6c5l5ygV45204Z6mRSDzcWosyOoz0w2difaA$>. You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[bfrosik]

That would be another issue. I can add an issue of creating an example beamline, very simple, with instructions how to modify it for the needs. Or would you like to add it?

[bfrosik]

Addressing the different types of shrink wrap, I believe we do not have to worry about number of parameters as long as we document it. There is a parameter in shrink wrap "type" that will define what parameters are needed. I can modify the GUI to change the layout with parameters depending on type. It was a TODO but as there wasn't a request, it was sitting on.

[jacione]

The gui already has a defaults button for most of the features.

I was thinking of this as being a little broader than just default values. There would be several presets for a given feature. Reusing the shrinkwrap example, you could have something like:

shrinkwrap_presets = {
  "aggressive": {
    "shrink_wrap_trigger": [1, 1],
    "shrink_wrap_type": "GAUSS",
    "shrink_wrap_threshold": 0.3,
    "shrink_wrap_gauss_sigma": 2.0
  },
  "tight": {
    "shrink_wrap_trigger": [1, 1],
    "shrink_wrap_type": "GAUSS",
    "shrink_wrap_threshold": 0.2,
    "shrink_wrap_gauss_sigma": 1.0
  },
  "default": {
    "shrink_wrap_trigger": [1, 1],
    "shrink_wrap_type": "GAUSS",
    "shrink_wrap_threshold": 0.1,
    "shrink_wrap_gauss_sigma": 1.0
  },
  "gentle": {
    "shrink_wrap_trigger": [2, 2],
    "shrink_wrap_type": "GAUSS",
    "shrink_wrap_threshold": 0.1,
    "shrink_wrap_gauss_sigma": 2.0
  },
  "permissive": {
    "shrink_wrap_trigger": [3, 3],
    "shrink_wrap_type": "GAUSS",
    "shrink_wrap_threshold": 0.05,
    "shrink_wrap_gauss_sigma": 2.5
  }
}

Maybe shrinkwrap is a bad example since the parameter space is relatively small. Some of the other features---GA and adaptive multipeak come to mind---have a very open-ended configuration space, much of which is largely useless. For example, you probably shouldn't cull all of your GA individuals on the first generation, but you could. That mistake is obvious to someone who knows how genetic algorithms work, but completely opaque otherwise.

My point is this: Cohere should be easy to use without understanding everything going on under the hood. And for the most part, I really think it is (especially compared to other BCDI packages). But I suspect that having some easy-to-understand presets would allow non-expert users to use the software more independently.