*spurperl has asked for the wisdom of the Perl Monks concerning the following question:*

Fellow monks,

Disclaimers:

* This is more a call-for-code-feedback than a question...

* Please mentally perform s/random/pseudo-random/ on this post

A weighted random generator is needed when events should happen at random, but some should have a higher probability of happening than others ("All animals are equal, but pigs are more equal than the others" ... Orwell). For example, I'd like to decide between 0, 1 and 2 randomly, with 0 and 1 having an equal chance to be chosen, and 2 having a chance that is twice higher.

The following code shows a function implementing this weighted random generator, with some test code. Please tell me what do you think about it... Any obvious slumps ? What about efficiency ?

Thanks in advance

Disclaimers:

* This is more a call-for-code-feedback than a question...

* Please mentally perform s/random/pseudo-random/ on this post

A weighted random generator is needed when events should happen at random, but some should have a higher probability of happening than others ("All animals are equal, but pigs are more equal than the others" ... Orwell). For example, I'd like to decide between 0, 1 and 2 randomly, with 0 and 1 having an equal chance to be chosen, and 2 having a chance that is twice higher.

The following code shows a function implementing this weighted random generator, with some test code. Please tell me what do you think about it... Any obvious slumps ? What about efficiency ?

Thanks in advance

#!/usr/local/bin/perl -w use strict; # Given an array of weights, chooses an array element # pseudo-randomly based on those weights. # # For example, given (1, 1.25, 3.6, 2) # The chance that the 2nd element will be chosen is # 3.6 times as large as the chance that the 0th element # will be chosen # sub choose_weighted { my @weights = @{$_[0]}; my $acc = 0; my @acc_arr; foreach (@weights) { $acc += $_; push(@acc_arr, $acc); } my $rand_val = $acc * rand; my $i = 0; ++$i while ($acc_arr[$i] <= $rand_val); return $i; } # Test code - just to prove that we get reasonable # distributions # # With the test array used below, $count[3] obviously # should be about twice as large as $count[1], etc... # my @ss = (1.75, 2, 3.6, 4); my @count = (0, 0, 0, 0); for (my $i = 0; $i < 500000; ++$i) { ++$count[choose_weighted(\@ss)]; } $, = "\n"; print @count;

Back to
Seekers of Perl Wisdom

Comment onWeighted random numbers generatorDownloadCode