Hi all - please fill out the poll to find a new time to meet: https://forms.gle/Up86DaSLd2cNjoAV8 Thanks to Anna, Alex, and Misha who have already filled it out.
@james.casaletto @yigitkucuk92 @jakubm @soroush.seylani @anuiris
1 Like
Hi @everyone - thanks to everyone who could make it to our new time: 11AM Pacific every other Monday. Here are notes from the meeting:
Anna Lewkowitz, @anna.lewkowicz Update on Lunar Regolith causal inference project
-
Started with the original images from the OSDR
-
Used Photoroom to segment into individual plant images
-
Used SOAPP to quantify shape and color, resulted in 2 files for each image: a file describing color, and a file describing shape
-
Discussion of whether the color quantification from the Canon camera is pre-filtered. Dr. Kris Peach has done some work on the color spectrum of these images and we could consult with her in future
-
Connected the RNAseq “GSM” number with the individual plant image files
-
Discussion of how to prepare the data for CRISP: the goal is to run CRISPv1.1 with the tabular RNAseq+SOAPP data; and CRISP3.0 with the raw images and RNAseq data; and compare.
Alex DongHyeon Seo, @alexdseo Potential new project, extension of GLARE
- Will discuss with Richard and come back with updates
- Can we do representation learning on OSD-476?
- Alex to present on the GLARE representation learning method in a next meeting. Future work will be to add causal aspect
Richard Barker @dr.richard.barker - New project idea!
ExoLab 11 timelapse of plants - on GitHub and being ingested into OSDR
Timelapse: https://www.youtube.com/watch?v=H1rKmOad3O4
- includes environmental parameters like humidity, CO2, temperature
- This data includes the moment when the leaves open and close, and also moments when env parameters goes up and down
- Which comes first? Which is causing the other? Can causal inference help us? - Yes!
- If anyone is interested in helping us get started with this new project, please DM @dr.richard.barker and/or @lauren.sanders !
4 Likes
Hi @everyone - apologies that I needed to change our time to 11:30AM Pacific today.
Alex @alexdseo reviewed a new project, incorporating causal reasoning with representation learning
- Submitted an abstract to ASGSR with Richard Barker and team
- GLARE: discovering hidden patterns in spaceflight transcriptome using representation learning
- representation learning = using a representation of the data for better performance of ML models, especially for high-dim or unstructured data
- example: autoencoders
- causal representation learning: the looking into the latent features for causality
- let’s compare causal representation learning to CRISPv1.1 and CRISPv3.0, in the lunar regolith project led by Anna! @anna.lewkowicz
For new members (@eman-abc and @shivamksharma202 and anyone else!) - a great starting point would be a literature review on causal representation learning - let me know if you have items you’d like to present in the next meeting!
2 Likes
Dear team @AIMLawg
We met yesterday 18.August and covered the following items:
- new member introductions
- summary of in silico causal inference methodologies / CRISP
- summary of OSDR data / portal
@anna.lewkowicz and @dr.richard.barker – hoping you can make the next meeting so we can discuss the lunar regolith project.
To the rest of the team, please feel free to explore OSDR for potential datasets as discussed on the call.
cheers
-james
8 Likes
Hi all, we had a short meeting today and reviewed the different causal inference algorithms and the goals of the group.
CRISPv3: GitHub - crowdplat/NASA-NOIS2-171
Yigit’s DoWhy implementation: GitHub - yigitk/crspv1.1-causal-inference: This is a Streamlit application for interactive causal analysis using the DoWhy library. It allows users to explore potential causal relationships in their data through an easy-to-use web interface.
@anna.lewkowicz @dr.richard.barker @yigitkucuk92 @james.casaletto @alexdseo please feel free to post here if you are running into any issues or questions with your causal inference work, or if you have any interesting results to share!
1 Like
Hi all - I’m cancelling today’s meeting since some folks are on vacation. See you in 2 weeks!
1 Like
Notes from today’s causal inference subgroup meeting:
Anna @anna.lewkowicz
RR9:
- Planning to run CRISP on the RR9 data
- Trying to install CRISP v3
- WIll run the RR9 image data while Daniya runs the tabular
Plant image project:
- tried to run v1.1 on the results from SOAPP, didn’t really work bc there wasn’t enough data
- so James @james.casaletto will run on images with CRISPv3
Daniya @daniya.ses
- Working on the AI/ML TOPS training
- Becoming familiar with the RR9 dataset
- Next steps, run tabular RR9 data through CRISPv1.1
Anna and Daniya will meet to run CRISPv1.1 and CRISPv3.0 synchronously on the RR9 data
2 Likes
Dear all - we will cancel today’s meeting due to schedule conflicts. Please post any updates or questions here. See you in 2 weeks! @AIMLawg
4 Likes