# 1:1 Rachel

## Fri, Sept 23

Logan is magically working again?!

Fixed a couple things on the SciPy poster by reloading some existing models now that Logan is working

Want to throw out ideas of labels and instead find fission/fusion events directly in the data
- start with pairs of images, windows of images
- work up to Conv-LSTM
- pre-trained models (e.g., ResNet)
- **come up with some kind of "score" at each time point, regarding how much fission/fusion there is**
- mixture model (fission + fusion)

May need a conversation with the Bio folks to make more concrete vocabulary

---

## Fri, Sept 9

https://github.com/rmattson1008/ornet

Check out Google Colab's GPU

Logan's GPU still down

"Other than that I really want to move forward with ideas about time series analysis. I'm having a good time reading papers and trying to make everything I read magically apply to ornet. Honestly my favorite part of the whole research thing."

Post ticket with Logan error message (or way of replicating)

---

Other date??

Taking 1 course this semester, 1 course next semester

Kinda resigned to continuing the project ???

Following up on temporal anomaly detection stuff
- wasting richness of data! classification based on a single frame doesn't make sense

compute
environment setup

ran with logan for the baselines
- ran into some issues

This week:
- clean up meekail's fork of ornet (that rachel has been working on)

Some kind of screenshare walkthrough of submitting GPU jobs, killing them, etc