i’ve spent so much of my life deciding what to do by evaluating tradeoffs - job alignment vs salary, working all the time vs maintaining friendships, being effective vs being silly, exploring technical interests vs other interests like health and clothes. only recently have i realized that, actually, tradeoffs are lame and none of this makes any sense
[...]
everything which is truly important will stick, and everything else was probably not as important as you thought, and optimizing tradeoffs is satisfying because it lets you feel like you’re being principled and rational but it is actually not an effective way to explore or resolve highly open-ended problems like “what should i do with my life”
vincent huang – YOU CAN GET WHAT YOU WANT OR YOU CAN JUST GET OLD
it's quiz day again. less anxious than before. probably because grades aren't that important anymore. and i know what to expect, and i have a cheatsheet. i remember my first quiz for this program, and how i felt after getting the auto-grading. it felt like the end of the world.
my phone has been on assistive access, there are so many inconveniences (no screenshot, no notifications, no zooming on camera) but its been great at making me use my phone less, i wonder how long i can last.
practicum meet n greet day where everyone meets some of the partners and ask questions. spent the rest of the day after quiz preparing good questions.
spent ~two hours reading papers for Meta on HDD failures and bit flip errors. it's fun reading papers, especially when its a new field, it's like discovering a new land that you've never ventured before, and trying to piece things together, why things exist, what problem it is solving, what the solution was, the results, and next steps and improvements. implementing papers would be even more fun, which is what will happen if i get into meta.
- Hard Disk Drive Failure Analysis and Prediction: An Industry View - Meta Research
- Evaluating and Enhancing Robustness of Deep Recommendation Systems Against Hardware Errors
- An Assessment of Vulnerability of Hardware Neural Networks to Dynamic Voltage and Temperature Variations
spent most of the time chatting with kevala, some links i found for probabilistic time series forecasting
- Quantile Time Series Regression Models Revisited
- Load probability density forecasting by transforming and combining quantile forecasts - ScienceDirect
- AutoBNN: Probabilistic time series forecasting with compositional bayesian neural networks
- Probabilistic Time Series Forecasting with Structured Shape and Temporal Diversity
both got me excited about the possibility of working with great people on interesting problems. there's so many rabbit holes to dive into, models to train, data to clean, papers to read, code to write. lets hope things go well. onto the interview grind.