I was in charge in testing a backend software that used a randomized algorithm. Different code path could be taken during different test runs. One path that was bad in particular. 😆To reproduce the bad path, I put a seed in the random number generator. Oh what fun testing randomized algorithm!
That's true, testing a randomized algorithm tends to be a lot trickier, and the shipping code as well as the test code should allow injecting and recording the random seed!
Man, this stuff is so fascinating to me - more so than the technical aspects of software development. The issues you're talking about on how to pick the right heuristic have everything to do with your company, the examples you're considering, your ability to measure certain things and not others, etc etc. There's just so much there that is really not that technical, yet impacts these giant, business critical decisions. The art of software!
I think it's technical, but it doesn't LOOK technical. IMO one of the biggest challenges we have as software engineers is really understanding what our customers are looking for, and figuring out ways to measure how well we're delivering it!
Thanks for the video! Slight off topic, but I've seen several examples where instead of one large optimization it was faster, simpler and of better quality to several times start with a random solution and do a local simple optimization then pick the best out of these several simple solutions.
I was in charge in testing a backend software that used a randomized algorithm. Different code path could be taken during different test runs. One path that was bad in particular. 😆To reproduce the bad path, I put a seed in the random number generator. Oh what fun testing randomized algorithm!
That's true, testing a randomized algorithm tends to be a lot trickier, and the shipping code as well as the test code should allow injecting and recording the random seed!
Man, this stuff is so fascinating to me - more so than the technical aspects of software development. The issues you're talking about on how to pick the right heuristic have everything to do with your company, the examples you're considering, your ability to measure certain things and not others, etc etc. There's just so much there that is really not that technical, yet impacts these giant, business critical decisions. The art of software!
I think it's technical, but it doesn't LOOK technical. IMO one of the biggest challenges we have as software engineers is really understanding what our customers are looking for, and figuring out ways to measure how well we're delivering it!
Thanks for the video! Slight off topic, but I've seen several examples where instead of one large optimization it was faster, simpler and of better quality to several times start with a random solution and do a local simple optimization then pick the best out of these several simple solutions.
That doesn't sound off topic at all, it sounds like a smart, practical approach. Thanks for posting!
Loved the video :) thanks for the insights
Thanks!
Wow, thank you! Completely unnecessary but appreciated!