search

credo-science / credo-webapp

Credo web application
MIT License
2 stars 6 forks source link

updating the static web-based filter #35

Closed piotrhomola closed 6 years ago

piotrhomola commented 6 years ago

The next step on the way to increase the clarity of the web-app, after enabling on-time visibility and providing limited-access "contest view", is to update the static web-based filter, according to the current status of our understanding the current images. I propose the following filter: 1) total on-time > 0, and 2) image taken after on-time incremented for the first time, and 3) avbrightness_max =< 0.01, and 4) maxbrightness_min >= 120, and

We show only statistics of users/teams that meet condition 1) and for those users/teams we display only images that meet 1), 2), and 3). Other images should be saved but not displayed and not counted for statistics. So in practice I think we do filtering for all images just once, assign a "display flag" Yes or No to each of them according to the filtering scheme, and later, continuously, to each newly received image, upon arrival.

The selection rules should be clear and well visible to the users, so I suggest adding a corresponding static sub-page named "help" or "FAQ" or similar, where we could have some universal text explaining the filters, and later maybe other essential info. What do you think?

koziomek commented 6 years ago

After filtering all detectiond through 3 and 4 we are now down to a little over 22k visible detections. Implementing 1 and 2 would require us to keep some state which we already agreed not to do.

piotrhomola commented 6 years ago

I do not quite see why 1 and 2 is not possible if 3 and 4 is possible. Please consider e.g. the following proposal: to determine 1 and 2, for each existing user you do once what you already do: check his ontime, then if it is > 0, you do one more check whether the user has any hits after first positive delta time had been received (after his historical start with positive ontime). If yes, then this particular user has event statistics > 0, and for all is events received after his historical start you can reduce 1 and 2 to another condition: if event statistics > 0 then you assign display Y to the image. Then you only need to treat the "first event" case: when event statistics is = 0 and an image is received, you request pings backwards, until you get positive delta time. If you do not get any positive delta time you have to assign display N, it might be a detection made with an older version of the app. Everything would be easier if we could afford a new column with "historical start", but I understand that we want to avoid new columns. What do you think?

No problem with a significant reduction of events after applying 3 and 4

W dniu środa, 27 czerwca 2018 Krzysztof Oziomek notifications@github.com napisał(a):

After filtering all detectiond through 3 and 4 we are now down to a little over 22k visible detections. Implementing 1 and 2 would require us to keep some state which we already agreed not to do.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/credo-science/credo-webapp/issues/35#issuecomment-400801672, or mute the thread https://github.com/notifications/unsubscribe-auth/Ahue6Zro1e9ZPPoS4LLlcqPmN2VLR7nLks5uA90zgaJpZM4U5gBs .

piotrhomola commented 6 years ago

Important thing: please do not apply any filters before they are not explicitly mentioned in the help/FAQ section, we do not want any cryptic re-ranking, everything transparent, users should have the impression that we know what we are doing :)

koziomek commented 6 years ago

Implementing 1 and 2 would require us to fetch additional information from the databse increasing the number of queries at least a few times. It would also make distributing the job of refiltering detections uch more diffucult and storage database heavy. The filters 3 and 4 only have to look at single detection to determine the result.

koziomek commented 6 years ago

There is no help/FAQ section on https://api.credo.science site, it is available at https://credo.science and that site is not in this repository.

piotrhomola commented 6 years ago

Sorry but I do not get your remark to implementing 1 and 2, could you please refer somehow to my proposal? Could you explain "Fetch additional information"? My proposal was about doing this only once, then we need only to modify one query by adding one more condition: number of events recorded so far. If we cannot restrict the view only to events taken during the detection with ontime then there is no sense in showing ontime, and if we can't show ontime there is no sense in showing anything... Please try to keep in mind the broader view of the mission, not only the technical point, and please do not hesitate to ask if in doubts, here, by e-mail, skype, phone, anyway - we need to understand each other. Thanks!

piotrhomola commented 6 years ago

Re FAQ: again it is about the choice between being user friendly and unfriendly. If we show something and not all the events are shown we need to explain at the place where we show, if we are friendly. So please consider addig the help section on api.credo.science where we would place only the text referring directly to the view we propose. Filtering information is essential to understand the view and it should be delvered clearly. If you refuse adding help then please remove all the filters. Thanks

piotrhomola commented 6 years ago

It seems I have to further clarify and strenghten my point by withdrawing my issue: if we cannot have easily 1 and 2 and 3 and 4, with a way to adjust 3 and 4 from time to time after a justified request, then please bring back the previous filter view asap, and I propose that we meet in person and discuss before taking further steps. Doing the new filter only partly, e.g. only 3 and 4 as you did - without consulting and with no explanation provided to the users on site, is in my opinion harmful to the project PR. Thank you in advance for your cooperation and understanding.

koziomek commented 6 years ago

I agree, we need a meeting in person to discuss some things. I don't see why changng the filter parameters now is different than before (example: when we first started filtering the data or when we were asked to hide all detections that are not from APIv2). I would also appreciate if you wrote what exactly should be in the FAQ.

piotrhomola commented 6 years ago

Ok, please bring back the old view and let's meet to discuss further things, let us set the date via e-mail.

W dniu czwartek, 28 czerwca 2018 Krzysztof Oziomek notifications@github.com napisał(a):

I agree, we need a meeting in person to discuss some things. I don't see why changng the filter parameters now is different than before (example: when we first started filtering the data or when we were asked to hide all detections that are not from APIv2). I would also appreciate if you wrote what exactly should be in the FAQ.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/credo-science/credo-webapp/issues/35#issuecomment-401097294, or mute the thread https://github.com/notifications/unsubscribe-auth/Ahue6fZPiSm9LNWyqaw0bM-XTv5FhY4sks5uBQZogaJpZM4U5gBs .

koziomek commented 6 years ago

We are waiting for the FAQ text.

piotrhomola commented 6 years ago

I propose the following:

API Frequently Asked Questions

  1. Which filters are applied on the current view and statistics?

<here your technical explanation, I can smooth the language if needed>

  1. Can I access the full data set, also the events filtered out for the web view? How?

Yes, the data access procedure is described here. <link on 'here' do the procedure described on github>

Will it be fine?

2018-07-05 11:58 GMT+02:00 Krzysztof Oziomek notifications@github.com:

We are waiting for the FAQ text.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/credo-science/credo-webapp/issues/35#issuecomment-402671360, or mute the thread https://github.com/notifications/unsubscribe-auth/Ahue6V0tAgD5NMRsl-TpNyR02x4twEBPks5uDeNggaJpZM4U5gBs .

koziomek commented 6 years ago

We are now relabeling all detections and hiding the ones received before user started sending valid on_time data.

What about 3 and 4? Do you have an idea about what those parameters should be?

piotrhomola commented 6 years ago

Please disregard 3 and 4, apply only the old filters plus 1 and 2, plus explanation. The contest view told us that for 3 and 4 we need a deeper study plus an algorithm to remove images with hot pixels, so I withdraw my earlier request. So for the time being: old filters unchanged + ontime (1 and 2) + explanation what we are doing in FAQ. Thanks!

W dniu czwartek, 5 lipca 2018 Krzysztof Oziomek notifications@github.com napisał(a):

We are now relabeling all detections and hiding the ones received before user started sending valid on_time data.

What about 3 and 4? Do you have an idea about what those parameters should be?

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/credo-science/credo-webapp/issues/35#issuecomment-402732881, or mute the thread https://github.com/notifications/unsubscribe-auth/Ahue6ZrrB5h3nTK-I8_-jEW54gfHf0P1ks5uDhzggaJpZM4U5gBs .

piotrhomola commented 6 years ago

You should contact Piotr Homola (Piotr.Homola@ifj.edu.pl) and describe...

Please replace with:

You should contact CREDO Science Coordinator (credodetector@credo.science) and describe...

W dniu czwartek, 5 lipca 2018 Krzysztof Oziomek notifications@github.com napisał(a):

Closed #35 https://github.com/credo-science/credo-webapp/issues/35.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/credo-science/credo-webapp/issues/35#event-1718209745, or mute the thread https://github.com/notifications/unsubscribe-auth/Ahue6YUAhy1XbsGVXbo1JfLL_kfD3WCmks5uDiEegaJpZM4U5gBs .

piotrhomola commented 6 years ago

Is the filtering finished? Detections "before the ontime era" still visible, see Leszczur or nkg9 in the user listing.

W dniu czwartek, 5 lipca 2018 Krzysztof Oziomek notifications@github.com napisał(a):

Closed #35 https://github.com/credo-science/credo-webapp/issues/35.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/credo-science/credo-webapp/issues/35#event-1718209745, or mute the thread https://github.com/notifications/unsubscribe-auth/Ahue6YUAhy1XbsGVXbo1JfLL_kfD3WCmks5uDiEegaJpZM4U5gBs .