Hello! My name is Amitav!

Today I did little work. This is bad. I read a lot of random things. I practiced competitive programming, completing question 2 from the 2011 CCC Senior division and getting stuck on question 4 before spending some time learning about max flow graphs, though tomorrow I must continue learning about them further. I finished listening to David Senra’s podcast on Claude Shannon, and began listening to his podcast on Phil Knight (Nike man). In addition, I began listening to a podcast from Machine Learning Street Talk with Llion Jones and Luke Darlow on a stagnating machine learning research landscape. They also discussed their own model, Continuous Thought Machines, though I don’t have too many thoughts on it as I didn’t read the paper in much detail. I completed none of the tasks I set out for myself yesterday. It’s also late. One thing I will try for tomorrow is restricting the set of media I am allowed to consume. In particular, tomorrow I will only consume my work Slack, work email, school email, school brightspace page, and “The Death of Ivan Ilyich”, along with my music. Another thing I began tracking today was how much profanity I used, which came to 140, +/- 10-15. That includes also times I had profane thoughts, with the number of times I spoke what I consider curse words probably in the single digits. I am attempting to cut down my profanity usage as it’s unprofessional and I would not like to accidentally curse in front of coworkers. My goal by next Monday is to bring that number down to at most 70/day, with at most one spoken instance per day. I believe this may not be ambitious enough, as I suspect beginning tracking caused me to curse more out of awareness of my tracking. My tasks for tomorrow are:

  • Complete project demo for 8090 (URGENT)
  • Start longer running experiments for my QSD paper to allow the models to train to convergence
    • Modify hyperparameters to be an unreasonable number of epochs (>10000) with early stopping
    • Rent out two CPU pods
    • Generate the 5 qubit and 8 qubit dataset, one on each pod
    • Rent out four pods on RunPod,
    • SCP each dataset to two of the pods
    • SCP over the relevant models to each pod
    • Begin training the models