I guess I should have continued to hang out here…

@rjnw You said last Wed. that you’d try nb_simpbucket.hs again — results?

I was fixing some bugs in nb_simpbucket

as the output wasn’t matching haskell

It is now the same as haskell

so I will try running it for full today and get timings

okay the previous naive bayes takes 2.7 seconds and the new one takes .26 seconds

I still haven’t fully verfied with haskell

this is runtime with my compiler

also I brought the 4s update to 2.7s for previous version by fixing some of my optimizations which were not working earlier for naivebayes :smile:

for haskell with previous naivebayes file between every “maybe sample new topic” line there is around 8seconds of gap but I am not seeing any difference between the prog generated by running new hk-maple

oops no thats the new haskell code, I forgot ken already updated it

if that 8second gap is for 100 updates then I am still way too slow compared to haskell

2.7 -> .26 is good!

I am not enough in the loop to understand what that ‘8 second gap’ refers to though.