Kathye - 2005-10-26

Really like Arboro, the first tree program I've felt I have a hope of understanding at any deep level.

I'm reading the Weber Penn article and wanted to ask about the section stating:

"To evenly distribute the splits, we use a technique similar to Floyd Steinberg Error diffusion. For each recursive level, a global value holds the 'error value' initialised to 0.0.

Each time nSegSplits is used, this error is added to create a SegSplits(effective) which is rounded to the nearest integer. The difference (SegSplits - nSegSplits effective) is subtracted from the error."

Is this necessary for determining useful figures to enter for nSegSplits in Arboro?

If so I don't understand fully how to derive an 'error value' and initialise it to 0.0

Any help appreciated.