Rescaling the Form-Factor Quality - #484
Conversation
| min_q_distance = 0.01 # Min distance btw peaks (in Q) | ||
| mqd_n = int(np.ceil(min_q_distance / delta_q)) # same in num frames | ||
| peak_ind = scipy.signal.find_peaks(-filtered, distance=mqd_n) | ||
| peak_prominence = (filtered.max() - filtered.min()) * 0.1 |
There was a problem hiding this comment.
How did you choose this criteria for peak prominence?
There was a problem hiding this comment.
Now I made it smaller. Such that to pass the tests. We may add some more tests, but these are quite aggressive already. I use more intensive noise in the tests that we really have.
| idxPlusErr = ffdata[:, 0].searchsorted(m1pos + maxXerr) | ||
| idxMinusErr = ffdata[:, 0].searchsorted(m1pos - maxXerr) | ||
| popt, pcov = scipy.optimize.curve_fit( | ||
| lambda x, a, b, c: a * x**2 + b * x + c, |
There was a problem hiding this comment.
Did you visually check any of the fits? Or test other function forms? Does it make a difference at this region of fitting?
There was a problem hiding this comment.
Fits are almost the same. Square function fitting of near-minimum vicinity is done to estimate position uncertainty from data uncertainty.
mdondrup
left a comment
mdondrup
left a comment
There was a problem hiding this comment.
I cannot assess the validity of the formula.
|
Does this affect the ranking of the quality score? From better < to better > ? |
|
Yes. The idea is to make FF quality working in the same way as OP quality. It will be bounded in (0,1) and oriented in the same way.
Best regards, Alex
Sent from [Proton Mail](https://proton.me/mail/home) for Android.
…-------- Original Message --------
On Saturday, 04/11/26 at 10:34 Michael Dondrup ***@***.***> wrote:
mdondrup left a comment [(NMRLipids/FAIRMD_lipids#484)](#484 (comment))
Does this affect the ranking of the quality score? From better < to better > ?
—
Reply to this email directly, [view it on GitHub](#484 (comment)), or [unsubscribe](https://github.com/notifications/unsubscribe-auth/AGBXJOECU4ED7WYQTNWDJ7L4VH7QTAVCNFSM6AAAAACXOAHQ6KVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHM2DEMRZGEYTINBUGE).
You are receiving this because you authored the thread.Message ID: ***@***.***>
|
|
I hope we can merge this soon @comcon1 . Once it is merged, I need to change the 'Best' ranking in the UI. Also, we need to bump python module in pip. |
@hsantila checked me and had some questions. I found a mistake while answering, and sent again. She checked again (not here), I answered (not here). Now waiting for approval. |
3 participants
FormFactor will be rescaled in this PR as we already discussed between collaborators. We will use the same probability estimate as for order parameters; it requires us to define the error of the position of the first minimum.
After it is done, we can issue a combined quality measure (will NOT BE DONE in this PR).
estimate_error_of_minestimate_error_of_minfmdl_evaluate_qualityAs a small useful bonus,
fmdl_evaluate_qualitynow supports--ff-onlyand--op-only. Default behavior (used in CI pipelines) was not changed.