Profiling this code shows the bulk of the time is spent on the log operation. Is there another way to write this in Python 3 for more efficiency? Replacing the loop with a list comprehension was actually less efficient and so was map because of lambdas.
def log_total(data):
total = 0.0
log = log(data)
for i in range(10000):
total += log/(i+1)
return total
Thanks!
I'd factor the log out of your summation and cache your sum:
harmonic_series = sum(1. / i for i in range(1, 10001)) # Thanks, @mgilson
def log_total(data):
return log(data) * harmonic_series
You could also use PyPy to speed it up even more.
log it would still be called 10,000 times thus incurring a huge cost each time, especially when you are essentially using a static variable, but computing it each call to the method. The next step would be to use memmoization as stated in the other question to get an even more speed from the function.
'(1./np.arange(1,10001)).sum()' cuts it down by a factor of ~6 for me, (but isn't as lazy)Can write with lambda in on line, like this:
total = lambda data: log(data) * sum(1.0 / i for i in xrange(1, 10001))
I used Python 2.7.3.
10000 elements is created (try changing 10001 to 1000000001 and see what happens), which takes up a ton of RAM.
1.0/i or use from __future__ import division, else your sum will equate to 1+0+0+0+....
logfunction?for i in range(1,10001): total += log/i1.42 msecper loop to1.09 msecper loop. That's a pretty big improvement.logfunction the one from themathmodule? If so, I doubt you'll be able to improve on its performance directly (since it's a built-in function implemented in C). Iflogis something you've coded for yourself, then there's might be room for improvement, but you'll have to show its code before we can suggest anything!