Edge Pixel Classification Using Automatic Programming

  • Kristin Larsen Østfold University College
  • Lars Vidar Magnusson Østfold University College
  • Roland Olsson Østfold University College


We have considered edge detection as a classification problem, and we
have applied two popular machine learning techniques to the problem and
compared their best results to that of automatic programming. We show
that ADATE, our system for automatic programming, is capable of producing
solutions that are as good as, or better than, the best solutions generated by
two other machine learning techniques.
The results demonstrates the ability of the ADATE system to create
powerful heuristics to solve image analysis problems.