The Natural Laws of Paper Deadlines

  1. Yes, no one will read your code. But it’s worth taking the care to design your system, from architecture to logging, properly. Shortcuts compound.
  2. Keep a journal. You will encounter the same problems twice.
  3. Write down your methods, especially after you submit the paper. Who knows, it may get accepted, and you may have to run more experiments.
  4. Collect every statistic you can think of. If you don’t, you’ll have to rerun everything on deadline day.
  5. Never do experiments on your personal computer. Just don’t.

This is a working list.


Now read this

2017 in Books

Here are the books I read (or rather, remember reading) in 2017. I don’t think star and number scales are particularly useful, so I mark each book as one of the following: Not worth reading Worth reading Must read Historical Fiction Fall... Continue →