<html>
<head>
<style><!--
.hmmessage P
{
margin:0px;
padding:0px
}
body.hmmessage
{
font-size: 12pt;
font-family:Calibri
}
--></style></head>
<body class='hmmessage'><div dir='ltr'>Hi all,<br><br>I still try to find a better split algo for splitter. Up to now it<br>is kind of luck when it finds a good split.<br>I've invested many hours, but I still don't fully understand<br>under what circumstances splitter should stop searching<br>for a (better) solution. It is still not possible to search all <br>possible splits, so brute force search doesn't work.<br><br>I also want to implement some kind of performance test to<br>avoid the situation that I optimize the algo for some countries<br>and produce worse results for others.<br><br>If you observe bad results with splitter,<br>please always send the densities-out.txt file<br>and either the splitter.log or the options in use<br>because with these rather small files I can reproduce<br>the problem.<br><br>Gerd<br> </div></body>
</html>