Prevent overflow errors when multiplying and powering in Python

Question

How would you prevent the errors when calling function biased_random defined below and what are the limits for arguments scale and bias to hold for preventing problems with big or small numbers?

def biased_random(scale, bias):
  return random.random() ** bias * scale

>>> sum(biased_random(1000, 10) for x in range(100)) / 100
64.94178302276364

>>> sum(biased_random(1000, 100000) for x in range(100)) / 100
0.0

>>> sum(biased_random(1000, 0.002) for x in range(100)) / 100
998.0704866851909

This seems clearly defined here en.wikipedia.org/wiki/Double_precision_floating-point_format. This defines the limits on floating-point numbers. 10**308 is pretty clearly defined as the upper limit. Is this what you're looking for? What more do you need to know? — S.Lott
– S.Lott, Commented Feb 22, 2012 at 15:19

inspectorG4dget · Accepted Answer · 2012-02-22 15:06:20Z

1

I'd use sys.maxint to figure out what the overflow point is. Then divide or nth-root it and compare with the number that you have:

r = random.random()
if sys.maxint ** (1.0/bias) < r:
    print "overflow imminent"
elif sys.maxint/float(scale) < r ** bias:
    print "overflow imminent"
else:
    print "overflow unlikely. To infinity, and beyond..."

Hope this helps

answered Feb 22, 2012 at 15:06

inspectorG4dget

115k30 gold badges159 silver badges253 bronze badges

Sign up to request clarification or add additional context in comments.

3 Comments

inspectorG4dget Over a year ago

Figure out what the overflow on your machine is (should be sys.maxint). Then you can test away. What I've posted is an implementation of a theoretic discussion I had with someone, some time back

xralf Over a year ago

I'm testing it and seems that it works (in accordance to limit 10**308 as S.Lott commented), but wonder how you solved it in the theoretical discussion.

inspectorG4dget Over a year ago

It was algebra. What you want to do is test that x * c (where c is a constant) will not overflow. You know (or can easily find) the overflow value. Then you do overflow/c if x is bigger than this number, then you're headed for an overflow.

Collectives™ on Stack Overflow

Prevent overflow errors when multiplying and powering in Python

1 Answer 1

3 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

3 Comments

Your Answer

Sign up or log in

Post as a guest

Related