ridge

Terms from Artificial Intelligence: humans at the heart of algorithms

A ridge in the fitness landscape is when there is a direction of change in the parameters that makes very little difference to the fitness function, but where small changes in any other direction drop the fitness radically. For example, suppose the fitness in an area is approximately 1−√| x−y | and we start at the (x,y) point (2,2) where the fitness function is zero. Small changes to (2+δx, 2+δy) stay at one if δxy, but if they are not equal even quite small changes where δx≠δy will lead to rapid drops in the fitness function;

Used on page 76