@nstoffle Thanks for sharing these HERA interactive plots! The ability to use the select tool and toggle the ascending/descending passes makes the SAA boundaries much easier to interpret visually.
I’m downloading the HTML files now to take a closer look at how the data filtering is structured under the hood. Really appreciate you sharing this baseline data with the group!
I was unable to write a proposal for the project. I was too busy in last month. When will the second phase start? @martin !
OK, no problem.
The second phase of the brainwriting exercise focused on commenting and reviewing the submitted proposals is planned to start next Tuesday during the joint AI/ML AWG meeting.
If anyone still intends to submit a proposal in the “brainstorming phase”, please do so as soon as possible. Use the original document template and keep the formatting unchanged. Copy paragraph 0 to the end of the document. Any formatting modifications create additional overhead when consolidating the inputs.
Once your draft is ready, send a private message to @martin and @odylec so everything can be prepared in time for Tuesday. @odylec supports the project as PM and takes care of the brainwriting process. If anything about the formatting is unclear, contact us directly.
Thank you.
Today we will begin the second phase of the brainwriting process. Please provide feedback in the form of comments,
and if you want to add anything to the text, do so only using Suggesting mode.
The commenting phase will run for 14 days.
This is also a reminder that the monthly AI/ML AWG meeting takes place today.
Martin
I noticed the document does not have any comments yet. Were they removed, or is this something from my side?
No comments have been added yet, as far as I can see. There is still time to do so until the end of February.
I have added one comment to verify that comments are visible. Could you please check whether you can see it? It would also help if you could try adding a comment yourself, just to confirm everything works correctly.
Please also verify that there are no permission issues with the document.
Yes, I was able to see the comment. The permissions are also good. I am going to go in and add a comment too.
Please let me know if there is anything else I can do.
The commenting phase did not work very well. Apart from one response, no comments were added to the submitted proposals, so I suggest we move forward in a different way.
I have added a table at the end of the document. Please, if you are the PI of one of the proposals, fill in your name there, since the proposals have been anonymous so far. If you would like to join a proposal and contribute to it, please also add your name to the table.
Please do this by the end of this calendar week.
Thank you.
@martin I have a new method. Am I able to add in ?
OK, please add it.
We already have reference data together with SAA passes labeled using the NASA SRAG methodology here:
https://colab.research.google.com/drive/1W-r58FOB6jp8Cm0wbB_ue6z1sskhJZUg
This means that the individual proposed methods can now be implemented, and each resulting selection can be compared with the spectrum produced by the NASA SRAG approach.
I would also suggest creating a second reference dataset based on manual labeling of SAA passes. For that, we would need a semi-automated annotation tool.
Would anyone be willing to develop such a tool?
Once such a tool exists, we can divide the sample dataset among volunteers and label it manually. We would then have two datasets for testing the models:
one based on the NASA SRAG methodology,
one based on manual labeling by volunteers.
So the current task I would propose is: to create a semi-automated tool for labeling SAA passes.
Hello Mr. Martin,
Sorry for the late response. I am currently working on developing a tool for this task. Just to be more specific, are you looking for a tool that provides the AI response and then can be validated by a human, or are you looking for a tool where the human does the labeling entirely? @martin
Hi Vineeth,
thank you for taking this on!
At this stage, we are looking for a tool where humans do the labeling entirely. The volunteers would use it to annotate a portion of the data, which will then serve as a training dataset. Different models will be trained on this and evaluated separately on a held-out portion of the data.
If anything is unclear, don’t hesitate to reach out.
Thanks again for your time,
Hello! I joined yesterday itself and I might be late for your current work, but I want to contribute but I am unsure what to do right now, if there’s anything I can do, I would really like to get involved based on your previous conversation I assume you are working on a tool i would like to contribute to that task as well if it is feasible
Hi All,
Here is a quick update on the current project status.
Project overview
A presentation describing the problem we are solving is available here:
It would be great if someone could prepare a one- or two-slide summary for newcomers. If you create such a slide, please let me know.
Brainwriting document
Proposed solutions are collected in the brainwriting document below. Please note that the author table at the end still needs to be filled in - add your name if you contributed a proposal or plan to help in a given direction. You are also welcome to comment on existing proposals or add new ones directly in the document:
SAA pass labelling app
@VineethErram is currently building an application through which volunteers will be able to label SAA passes. I will announce here once the app is ready. To eliminate systematic errors, we need several volunteers to participate in the labelling. The training dataset covers approximately 100 days - with 10 volunteers, each person would need to label around 10 days. Full coverage of the dataset is not required.
The sample data are available here:
Volunteers, please sign up in the table here:
Once the manual labeling application and manual labels are ready, we will move on to a competition between individual ML models.
Newcomers
If you are joining the team, please register your contact details and Google Drive account here:
https://docs.google.com/document/d/1Faiq4lZarZ6_XKOEprOfBbB4o3K7gny8usn63r2KcJ4/edit?usp=sharing
Thank you!
Martin
Final Reminder – SAA Manual Labeling Project
Hi everyone,
This is a friendly reminder that the deadline for the SAA Manual Labeling Project is 4 August 2026.
If you have already completed your assigned day, thank you very much for your valuable contribution! If not, there is still time to complete your labeling before the deadline.
Please remember to:
Resources:
Thank you all for your time and support. Your contribution is helping us build a high-quality ground-truth dataset for the evaluation of future ML models.
If you have any questions or encounter any issues, please feel free to reach out.
Best,
Odysseas
Hi @martin and everyone! I’m a new AI/ML AWG member and an undergraduate physics student at Caltech working on machine learning for CMS particle-physics data.
I’m very interested in the SAA and GCR separation project because it connects closely with my background in particle physics, detector simulation, and classification. I’ve added my information to the newcomer document and will claim an available day for the current manual-labeling effort before the August 4 deadline.
I would also love to contribute when the project moves into the ML-model comparison phase. Please let me know if there is anything else I should review or prepare in the meantime.
Hi @odylec ! I claimed the 15 Dec–25 Dec range in the volunteer tracker, but the resource names in the reminder do not appear as hyperlinks for me. Could you please share the direct links to the Volunteer Guide, SAA Manual Labeling Tool, and Volunteer CSV Folder?
Could you also confirm whether I should produce one CSV for each date in the assigned range or one file for the full range? Thank you!
