Markdown "diary" of daily research efforts. Includes links to plots, results, outside posts/sites.
Rules:
- 10 mins is more than enough
- Set goals in morning, report at end of day
- Ask Braxton to approve PR
Plan:
- Code for discrete chebyshevs, with unit tests
- Preliminary code for cluster function averages over structures
- Read 30 mins, make notes
- Work on 318 or 601 a bit
- Anything on NNs or image AI
- No checking email between 7:15 am and 4:30 pm
- Stretch twice
- submit receipts ✔️ Respond to Vania
Plan: ✔️ Bike ✔️ Email Tyler about what I learned yesterday.
- Read 30 mins, make notes
- Work on 318 or 601 a bit
- Anything on NNs or image AI ✔️ No checking email before 4:30 ✔️ Stretch twice
- submit receipts
Spent hours learning about discrete orthogonal polynomials and their intervals. Would like to redo CE with a different choice. Saved some papers to "Cluster Expansion" in zotero.
Plan: ✔️ Meet with Mark, Tyler, Jared about DD paper ✔️ Work on DD paper
- Vania email/doc
- Work on 318 or 601 a bit
- Anything on NNs or image AI ✔️ No checking email before 4:30
- Stretch twice ✔️
- submit receipts
Focused exclusively on the invariant polynomial stuff for CE today. Made some great connections between the chebyschev stuff and the invariant poly stuff. Took good notes in my rocketbook.
Plan: ✔️ Bike + strength training (fastest bike ever, 8 chinups maxxed) ✔️ Talk Tyler about invariant polynomials, making basis orthornormal, restricting the field
- Read 30 mins, make notes
- Work on 318 or 601 a bit
- Anything on NNs or image AI
- No checking email before 4:30
- Stretch twice
- submit reciepts
I spent most of my discretionary time talking to Tyler and working on an example.
Plan: ✔️ Talk/email Tyler/Rod about SDG and Chap 7
- Read 30 mins, make notes
✔️ Work on 318 or 601 a bit - Anything on NNs or image AI
- No checking email before 4:30 ✔️ Stretch twice (did it three times)
Plan: ✔️ Bike + strength training ✔️ Talk/email Tyler/Rod about SDG and Chap 7
- Read 30 mins, make notes
- Work on 318 or 601 a bit
- Anything on NNs or image AI
- No checking email before 4:30 (didn't do this well, but better than average)
- Stretch twice (only did it once) ✔️ Feedback on the Oliver paper. Argh. ✔️ Get housing for NSF trip
Reached out to Alejandro for some insight on the cluster enumeration problem.
Getting the feedback on the paper to Oliver was critical and time consuming. And the CE reading/thinking was important.
A pretty good day for productivity.
Plan:
- Talk/email Tyler about SDG and Chap 7
✔️ Get MP and autoGR data and compare - Read 30 mins, make notes
✔️ Work on 318 a bit (worked on this A LOT) - Anything on NNs or image AI
✔️ No checking email before 4:30
✔️ Stretch twice
✔️ Anki review
Daily notes:
The code for doing the double descent plots with Mark are in the directory projects/double_descent. The data for the urgr and urmp energies are in juliaCodes/ClusterEnumeration.jl/data/. Code used to do today's work on comparing data from two different kpoint meshes and CEs for each dataset: makeCE.jl.
The difference between autoGR and MP energies is really small. The ~4 meV outlier seems to be due to a really uneven kpoint spacing for the MP grid, much less dense in one direction.
The fitting errors on both sets of data are also small.
✔️ Bike + strength training
✔️ Get new ID card
✔️ Meet with Mark, Jared, Tyler
🔲 Talk/email Tyler about SDG and Chap 7
🔲 Get MP and autoGR data and compare (tried to do this---kept getting interrupted)
🔲 Read 30 mins, make notes
✔️ Work on 318 a bit
🔲 Anything on NNs or image AI
✔️ Talk to Eric about 601 TA
✔️ Stretch twice (only did it once)
✔️ One extra positive interaction
Lost the whole afternoon to talking with potential students. Could have done more if that hadn't happened. Stayed off email most of the day, but didn't do it at the end of the day as planned...