Neuro-Hill-Climber: A New Approach Towards More Intelligent Search and Optimization

The paper proposes a hybrid mutation based search method, namely the neuro-hill-climber algorithm (NHCA), which can learn the problem structure on-the-fly and is able to deliver it in a comprehensible form to human researchers. The NHCA employs an Associative Artificial Neural Network to extract the intrinsic problem anatomy from search experience. The advantage of the proposed approach is that knowledge is derived from already converged states, leading to a more exact model of the problem structure. Thus, the proposed method may be contributory to many research areas where an accurate insight in the hidden structure of a problem can lead to better understanding and advance in theoretical knowledge.