Salting Meeting Minutes
===
- **Location:** https://ucdenver.zoom.us/j/97994358165
- **Date:** Nov 7, 2024, 8 AM Mountain
- **Meeting Notes**
- Mithun is presenting on possible timeGAN projects
- He points out the lack of diversity in generated pulses relative to the input data
- Amy: make a list of expected features in generated pulses
- Mithun would like to be done with TimeGAN work in the first quarter of 2025, applying to PhD programs in September 2025

###### tags: `Salting` `Meeting`
- **Location:** https://ucdenver.zoom.us/j/97994358165
- **Date:** Sept 5, 2024, 8 AM Mountain
- **Meeting Notes**
- Ben:
- Using "pre-DC3" simulated data, 10,000 events, monoenergetic, HV
- Trying to get better distribution for the z coordinate
- Attach a scalar quantity to the raw pulse like rise time
- Conclusion is that we can't get position information out of the simulation data, it just doesn't seem to have any differences in rise time, etc. that we see in real data
- Zack
-

:::info
- **Location:** https://ucdenver.zoom.us/j/97994358165
- **Date:** Aug 5, 2024, 9 AM Mountain
- **Meeting Notes**
- Simulation wishlist, https://confluence.slac.stanford.edu/display/CDMS/ML+position+reconstruction+-+wishlist
- The Toronto team is looking at using "Reduced Quantities", namely pulse amplitudes and time from trigger of the signal, to predict X, Y, and Z location
- Using 0V simulation data, 10 keV peak

:::info
- **Location:** https://ucdenver.zoom.us/j/97994358165
- **Date:** July 29, 2024, 9 AM Mountain
- **Meeting Notes**
- Report from Ben: training isn't trustworthy because it does not reproduce quantities that look like the quantities from histograms
- Sukee, Amy: in general we have the problem that our simulated data does not look like our real data, for example our real data has a "leading pulse" feature that is not captured in the simulated data
- Farnoush: https://arxiv.org/abs/2307.06855 may be a way to augment the data and generate more accurate fake data
- Next week the Toronto group will present on their work with position reconstruction

:::info
- **Location:** https://ucdenver.zoom.us/j/97994358165
- **Date:** June 24, 2024, 9 AM Mountain
- **Participants:**
- Amy Roberts (AR)
- Kitty Harris
- Steve Yellin
- Sukee
- Belina
- Janice
- Stephanie
- **Meeting Notes**
- Stephanie and Janice will give an update on their work next Monday
- Sukee will present on LSTM networks and interpreting classification the week afterj

:::info
- **Location:** https://ucdenver.zoom.us/j/97994358165
- **Date:** June 24, 2024, 9 AM Mountain
- **Participants:**
- Amy Roberts (AR)
- Pam Stark
- Kitty Harris
- Farnoush
- Zack
- Steve Yellin

- **Meeting Notes**
- Amy will get the pre-proposal for the workshop finished *this week*
- And start a list of people to invite
- Zack: what are these XOF and YOF quantities
- Quantities that roughly describe what sensor the event happened most closely

:::info
- **Location:** https://ucdenver.zoom.us/j/97994358165
- **Date:** June 24, 2024, 9 AM Mountain
- **Participants:**
- Amy Roberts (AR)
- Ben Lichtenberg
- Steve Yellin
- Farnoush
- Zack

- **Meeting Notes**
- **data for Chris**
- Chris may already have written a memorandum of understanding. Steve will take a look.
- Sukee will work on putting cuts into git-lfs
- **Ben's project**
- Ben and Sukee are working on documenting his projct on Confluence, https://confluence.slac.stanford.edu/display/CDMS/ML-based+position+reconstruction+in+SNOLAB+Detectors+using+raw+traces
- **Salting effort**
- We're meeting with the workflow specialist on Thursday
- Dennis is ramping up and will look at hyperparameters with an aim to speeding up the training
- Zack is training the network, it's going slowly but making progress. Probably will take about a week to train.
- re: predicting pulse shape based on location, Sukee suggests first concentrating on amplitude and then considering rise time
- **Sukee's efficiency cuts**
- Sukee is working on reproducing the processing for her RQ's, uses CDMSBats v5.6.1
- https://gitlab.com/supercdms/Reconstruction/cdmsbats/-/blob/develop/docs/installation_oldVersions.md?ref_type=heads

:::info
- **Location:** https://ucdenver.zoom.us/j/97994358165
- **Date:** May 13, 2024, 9 AM Mountain
- **Participants:**
- Amy Roberts (AR)
- Ben Lichtenberg
-

:::info
- **Location:** https://ucdenver.zoom.us/j/97994358165
- **Date:** May 6, 2024, 9 AM Mountain
- **Participants:**
- Amy Roberts (AR)
- Zack Kromer (ZR)
- Farnoush Banaei-Kashani
- Kitty Harris
- Sukee
- Steve Yellin (!)
- Chris Sanford
- **Meeting Goal(s)**
- Salting paper at https://www.overleaf.com/2478459487gfcbtzwhrtbj#5237f9
- Amy will reach out to Joel to see if he can come talk to us about the threshold settings and how the DAQ reacts.
- We need actual values for these settings!
- CDMSlite describes this in https://arxiv.org/pdf/1808.09098, see table 1 and figure 9 for threshold information
- How do we find this information for Kitty's UMN data?
- Chris Stanford is looking for raw data, including junk runs where the HV was turned off, really noisy data, ideally with human-generated labels like the data quality database and the ELog
- Sukee has used R135 and that entire R133-R135 dateset is on the OSN
- IOLibrary, pyRawIO, matCAP
- data quality database
- Are there ELogs? Joel, Miriam, Tina
- Summary documentation?
- https://scdms.slac.stanford.edu/cdms_restricted/dokuwiki/doku.php?id=analysis:r133:cdmslite:run3_summary
- https://scdms.slac.stanford.edu/cdms_restricted/dokuwiki/doku.php?id=analysis:r133:cdmslite:run2_summary
- Sukee is looking at the processed data with ROOT files
- Sukee is also interested in being able to process the data without MatCAP. Chris Stanford: we really just need the labels so probably a no on re-writing MatCAP
- Sukee recommends looking at 10 keV peak for understanding relationship between channels. She also recommends looking at SNOLAB detector data, which has improved simulation data.
- 10 keV peak in CDMSLite run 3 specifically

:::info
- **Location:** https://ucdenver.zoom.us/j/97994358165
- **Date:** April 22, 2024, 9 AM Mountain
- **Participants:**
- Amy Roberts (AR)
- Zack Kromer (ZR)
- Farnoush Banaei-Kashani
- Kitty Harris
- James Amidei
- Sukee
- **Meeting Goal(s)**
- Amy will outline the paper so we can look at it together next week and get on the same page
- Zack will present on the physics constraints, his understanding of them, and implementation ideas for them
- James Amidei presents his work on derivatives. See https://gitlab.com/supercdms/Analysis/photoneutron_multiple_scatters/-/blob/main/Pulse%20Derivatives%20/Pulse%20Analysis%201.ipynb?ref_type=heads,
https://gitlab.com/supercdms/Analysis/photoneutron_multiple_scatters/-/blob/main/Pulses.ipynb?ref_type=heads. Farnoush asks about how an "event" gets defined. Sukee explains that we have an activation and deactivation threshold. Often they're set at 5 sigma above the noise, but for e.g. R135 they're set lower(?). See https://arxiv.org/abs/1808.09098
- Amy will reach out to Joel to see if he can come talk to us about the threshold settings and how the DAQ reacts.
- Farnoush: Zack, could you focus on spatial-temporal localization and thresholds. Amy: figuring out location based on pulses is something Aditi has worked on, using NxM optimal filter. Belina's new master's student is looking at using raw traces to extract this information, as well.
- ACCESS allocation is still live!
- https://allocations.access-ci.org/project-types
- https://docs.google.com/document/d/1LMdHunvjN63gxNMkAqXl8Ne2djMN4BMByCyVZw2Y9oU/edit?usp=sharing

## Summary

###### tags: `Salting` `Meeting`
:::info
- **Location:** https://ucdenver.zoom.us/j/97994358165
- **Date:** April 15, 2024, 9 AM Mountain
- **Participants:**
- Amy Roberts (AR)
- Zack Kromer (ZR)
- Farnoush Banaei-Kashani
- Kitty Harris
- **Meeting Goal(s)**
- https://www.ncfdd.org/

## Summary

###### tags: `Salting` `Meeting`
:::info
- **Location:** https://ucdenver.zoom.us/j/97994358165
- **Date:** April 8, 2024, 9 AM Mountain
- **Participants:**
- Amy Roberts (AR)
- Zack Kromer (ZR)
- Farnoush Banaei-Kashani
- Kitty Harris
- **Meeting Goal(s)**

## Summary

:::info
- **Location:** https://ucdenver.zoom.us/j/97994358165
- **Date:** March 3, 2024, 9 AM Mountain
- **Participants:**
- Amy Roberts (AR)
- Zack Kromer (ZR)
- Farnoush Banaei-Kashani
- Kitty Harris
- **Meeting Goal(s)**

## Summary
- Amy's action items:
- get Zack what we've done before on modeling the signal
- talk with the simulation group to make sure she hasn't missed anything
- get Zack the WIMP limit ipynb
- ~~set up agenda for program officer visit - it'd be nice to hold a workshop!~~
- ~~talk with students about looking at the data to make sure the monotonic constraint is accurate~~ James will start looking at this!
- ~~Dear Prisca, if I don't hear back from you by the end of the week we're making the data public!~~ Prisca is on board with making the data public!

## Summary
<!-- Please fill me in after the meeting -->
- Kitty: the parameter-salt code is working. Will work on adding this to the data and then processing
- Zack: Snapshotting after a specified time won't take long.
- Farnoush/Zack: how long does paper reading take? It depends on if you're skimming or going deep into the paper! About one paper per day when you're reading at low resolution. So expecting about two weeks for the literature review.
- Amy/Farnoush: Farnoush's previous workshop was in April, gave people 6 months advance notice. Want to find the call first to see how much detail they want. Some of them ask for many more details than others.
- Reserve money for media/video recordings.
- Write a whitepaper? This is particularly nice when there are new things under discussion. For us we'd need a new topic since this has been done before.
- Do we want to organize the workshop to coincide with the data science symposium?

:::info
- **Location:** https://ucdenver.zoom.us/j/97994358165
- **Date:** Feb 26, 2024, 9 AM Mountain
- **Participants:**
- Amy Roberts (AR)
- Zack Kromer (ZR)
- Farnoush Banaei-Kashani
- Kitty Harris
- **Meeting Goal(s)**

## Summary
<!-- Please fill me in after the meeting -->
- Kitty: the parameter-salt code is working. Will work on adding this to the data and then processing
- Zack: Snapshotting after a specified time won't take long.
- Farnoush/Zack: how long does paper reading take? It depends on if you're skimming or going deep into the paper! About one paper per day when you're reading at low resolution. So expecting about two weeks for the literature review.
- Amy/Farnoush: Farnoush's previous workshop was in April, gave people 6 months advance notice. Want to find the call first to see how much detail they want. Some of them ask for many more details than others.
- Reserve money for media/video recordings.
- Write a whitepaper? This is particularly nice when there are new things under discussion. For us we'd need a new topic since this has been done before.
- Do we want to organize the workshop to coincide with the data science symposium?

:::info
- **Location:** https://ucdenver.zoom.us/j/97994358165
- **Date:** Feb 19, 2024, 9 AM Mountain
- **Participants:**
- Amy Roberts (AR)
- Zack Kromer (ZR)
- **Meeting Goal(s)**

:::

## Summary
<!-- Please fill me in after the meeting -->

## Notes
<!-- Other important details discussed during the meeting can be entered here. -->
- Amy's action items:
- get Zack what we've done before on modeling the signal
- talk with the simulation group to make sure she hasn't missed anything
- set up agenda for program officer visit - it'd be nice to hold a workshop!
- ~~talk with students about looking at the data to make sure the monotonic constraint is accurate~~ James will start looking at this!
- get Zack the WIMP limit ipynb
- ~~Dear Prisca, if I don't hear back from you by the end of the week we're making the data public!~~ Prisca is on board with making the data public!