So hakrit is (now) way faster and significantly more accurate?

What kind of accuracy does one normally expect for NaiveBayes?

also I don’t know if this information is usefull but hakrit accuracy doesn’t change much after the first sweep, and hakaru’s also remain in the same range hakrit std-dev 0.33 hakaru std-dev 2.4

@rjnw I’m confused about the chart — why doesn’t the hakrit line extend as far as the haskell line?

so the x-axis is time in seconds, and total points I have are ten sweeps

hakrit takes very less time for 10 sweeps

and haskell takes a long time

I can extend hakrit line by doing more sweeps but then the number of sweeps won’t be the same

here is accuracy with the number of sweep

let me run it again I think haskell holdouts might be diffrent then ours