New Year Resolutions

3 minute read
Piled Higher and Deeper: https://phdcomics.com/comics.php?f=1670

Piled Higher and Deeper: https://phdcomics.com/comics.php?f=1670

Resolutions are meant to be broken, but like everything good in life, it’s better to try and fail than never to have attempted at all. So here’s my attempt for 2022.

1. Be a better modeler

2021 felt quite surreal — I got my first job offer as a machine learning data scientist, something I’d been dreaming of for nearly two years. In 2022, I wanna up my ante in the modeling game by working through (a subset of) the following:

  • Read Kevin Murphy’s Machine Learning: A Probabilistic Perspective to build a solid foundation πŸ‘‰ I learned ML in a somewhat hacky way, by doing projects across a wide range of topics (e.g., graphs, NLP, recommender systems, etc.). Sure, I’ve taken ML classes and read other ML books, but Kevin Murphy’s 1000-page bible is something I just got the courage for and probably closer to my training in computational cognitive science (i.e., structured probabilistic models and Bayesian stuff) than the other classics out there.
  • Apply best practices from Bayesian Modeling and Computation in Python to my dissertation πŸ‘‰ Much like how I learned ML for data science, I also picked up Bayesian modeling for research in a hacky way, by reading other scientists' code and asking for help, which is both a source of pride and insecurity. I’m thankful how this PyMC3 book finally came out when I’m wrapping up grad school and writing my dissertation — in the last 5 months of my PhD career, I want to re-learn Bayes the systematic way and double-check the modeling work I’ve done.

  • Learn more about deep learning by taking Hugging Face’s transformer and NYU’s DS-GA 1008 courses, listening to the Gradient Dissent podcast, playing with Papers with Code, and perhaps reaching out to ML Collective and/or Eric Jang.

I also wish to learn more about model serving — the reason being that deploying my own models may give me more autonomy and save loads of back-and-forth communication with machine learning engineers who may not be as familiar with my work.

2. Be a better friend

“It’s only love, it’s only pain, it’s only fear that runs through my veins, it’s all the things you can’t explain, that makes us human.” — Civil Twilight, Human (2009)

The pandemic still loomed over 2021 (and will stay for God knows how long…), yet I made more friends than I ever did, thanks to sports, tech, interviews, and circumstances stranger than Hollywood horrors. I’m so tried of telling the last story nor do I wish to anymore. The upshot is that I’ve felt pain, had fun, and seen the good, the bad, and the ugly in people — and I wouldn’t trade it for the world.

In 2022, I want to be kinder, more open, and let people feel valued and loved. As a researcher who works with kids, I used to end experiments with a generic “Thank you”, but now I always tell the family how their selfless contribution helps our lab carry on with cool science on rainy days — maybe I’m older, or maybe it’s the apocalypse, I’ve learned nothing is ever easy in this life and we can take no kindness for granted. And as an introvert, it used to kill me to reach out to folks I admire, but as I’m doing that increasingly more, I feel happier than ever — this may be the social version of fortuna favet fortibus (“fortune favors the bold”).

3. Be kind to myself

I’m glad I went to LA on a whim at the end of 2021 and had extensive conversations with my friend (who I bet my πŸ’° on will be a gifted clinical scientist) about mental health and what “doing enough” means. All this weight we carry and rituals we perform in order not to fall behind may just ironically weigh us down. I hope that in 2022, I can learn to let go of the shackles I put on myself and… just do stuff. Last but not least, I’ll shop around for a cat and a therapist I really like.