I m having a application in which i have to select a number out of many numbers according to their weights. Every time I will select , I have send the result to flash.I have found a algorithm in python. I have implemented it in php and was testing for its results. If i was running that algo in python it was giving good results but in php not so good. Ex. (1=>30,2=>40,3=>30) After running many times , the probablity of occurence first number in weighted array is always more but in python it is uniform. I have attatched the PHP code.
define("MAX",100000);
$reelfrequencies=array(30,40,30);
echo weightedselect($reelfrequencies);
/*function weightedselect($frequency)
{
$arr=cumWghtArray($frequency);//array(35,96,100);
print_r($arr);
$len=sizeof($frequency);
$count=array();
echo $r=mt_rand(0,$arr[$len-1]);
$index=binarysearch($arr,$r,0,$len-1);
return $index;
}*/
function cumWghtArray($arr)
{
$cumArr=array();
$cum=0;
$size=sizeof($arr);
for($i=0;$i<$size;$i++)
{
$cum+=$arr[$i];
array_push($cumArr,$cum);
}
return $cumArr;
}
function weightedselect($frequency)
{
$arr=cumWghtArray($frequency);//array(35,96,100);
$len=sizeof($frequency);
$count=array();
$count[0]=$count[1]=$count[2]=0;
for($i=0;$i<MAX;$i++)
{
$r=mt_rand(0,$arr[$len-1]);
$index=binarysearch($arr,$r,0,$len-1);
$count[$index]++;
}
for($i=0;$i<3;$i++)
{
$count[$i]/=MAX;
echo $i." ".$count[$i]."\n";
}
}
function binarySearch($ar,$value,$first,$last)
{
if($last<$first)
return -1;
$mid=intVal(($first+$last)/2);
$a=$ar[$mid];
if($a===$value)
return $mid;
if($a>$value&&(($mid-1>=0&&$ar[$mid-1]<$value)||$mid==0))
return $mid;
else if($a>$value)
$last=$mid-1;
else if($a<$value)
$first=$mid+1;
return binarySearch($ar,$value,$first,$last);
}
Here is the Python Code. I have taken this code from this forum . import random import bisect import collections
def cdf(weights):
total=sum(weights)
result=[]
cumsum=0
for w in weights:
cumsum+=w
result.append(cumsum/total)
return result
def choice(population,weights):
assert len(population) == len(weights)
cdf_vals=cdf(weights)
x=random.random()
idx=bisect.bisect(cdf_vals,x)
return population[idx]
weights=[0.30,0.40,0.30]
population="ABC"
counts={"A":0.0,"B":0.0,"C":0.0}
max=10000
for i in range(max):
c=choice(population,weights)
counts[c]=counts[c]+1
print(counts)
for k, v in counts.iteritems():
counts[k]=v/max
print(counts)
Problem is of mt_rand() function which is not uniform. The python random.rand() is very much uniform. Which random function should i implement in php with a proper seeding value every time it runs. I was thinking of using Withcmann (used by python random.random) but how will i provide the seed.
