Acceleration of EC convergence with landscape visualization and human intervention

We propose Visualized EC/IEC as an evolutionary computation (EC) and interactive EC (IEC) with visualizing individuals in a multi-dimensional searching space in a 2-D space. This visualization helps us envision the landscape of an n-D searching space, so that it is easier for us to join an EC search by indicating the possible global optimum estimated in the 2-D mapped space. We first compare four mapping methods from the points of view of computational time, convergence speed, and visual easiness to grasp whole EC landscape with five benchmark functions and 28 subjects. Then, we choose self-organizing maps for the projection of individuals onto a 2-D space and experimentally evaluate the effect of visualization using a benchmark function. The experimental result shows that the convergence speed of GA with human search on the Visualized space is at least five times faster than a conventional GA.