Re: Constraint finishing: "Ideal"

On 9/17/14 10:31 AM, Harald Alvestrand wrote:
> On 09/17/2014 03:48 PM, Jan-Ivar Bruaroey wrote:
>>
>> 1.
>>       *  change algorithm to have distance converge toward 1:
>>         |(actual - ideal)/max(|ideal|,|actual|)|
>>
>
> That actually solves the zero problem too... except for the "ideal =  
> -5, actual = 5" case, which would still give 2, not 1. An argument for 
> not tolerating negative numbers.

But it's never larger than 2. If the ideal is x and you give me -x, 2 
actually strikes me as a reasonable penalty distance - not only is it 
far away, it is in the opposite direction as well - Ruling out negative 
numbers forever on the other hand seems overly restrictive.

.: Jan-Ivar :.