Last night I was faced with a daunting task of making a meta-heuristic process a non-trivial amount of data in no more than a tenth of the time it was taking now.

Just to help you embrace the herculean task this was: My goal was to make the algorithm’s runtime lower than 1.5 seconds for any realistic data sample. It was taking anywhere from 5 to 20 and even 30 seconds at the time.

So I set off. First some profiling and some tidbits here and there, then tweaking and poking and yanking and scraping code for hours upon hours upon hours.

Nothing, no increase in performance. Nothing even remote to sane behaviour at all. Even though the algorithm was working, my debugging/profiling data wasn’t making any sense.

And I said fuck this shit and went to chillax and rest my mind over Men Who Stare at Goats for two hours.

There were some good laughs and some pretty good chillaxing being done. My I-don’t-know-which-exactly-or-what’s-it-called part of the brain was probably busy processing the afore mentioned problem in the background. Or maybe it wasn’t, who knows.

All I know is that when I came back to the problem I noticed something funky.

HEY THAT FUNCTION ARGUMENT SHOULDN’T BE A FUCKING CONSTANT WHAT THE FUCK IS GOING ON!

Characters 1 and t look a lot alike …

So anyway, long story short. The meta-heuristic that used to take 1500 epochs to do something now takes only ten to do the same thing, but slightly better.

That’s a pretty fucking good improovement right there!

Now the algorithm is constantly taking about a second to do its thing and I’m happy. Pushed it into production where profiling data will be collected of a more varied sample of data to see if all is well.

Chillaxing! It really does work.

Tags: Algorithm, Brain, food for thought, life