if your problem is interesting, you’ll be helped. if your disarray would be interesting, you will be thrown into chaos. either way, if it’s interesting, something will be changed
loss functions and training setups >>> techniques and rules
do it either right now or probably never.
you see what you aim at; you aim at what you see.
programming is easy u just open up the magic computer box & make a wish
The usefwl parts of a math model are usually those which most readily convert to common tongue. Beware Fisherian-runaway abstractions, and don't settle for mere hyper-specific insights. Ground yourself in what applies to the contexts you terminally care about.
How could you have thought this faster?
If what you're saying is true, it should be convergently true and there should be multiple roads to understanding it from the available evidence.
Skill comes from making smaller circles.