Selection for Reinforcement-Free Learning Ability as an Organizing Factor in the Evolution of Cognition

Figure 10

Idealization of the solution evolved in runs 0, 1, 2, 3, 5, 7, and 8 of the main model. These patterns were named L (for left turn), R (for right turn) and SJ (for step or Jump). Signs of the L pattern are flipped with respect to Figure 9. L and R detect horizontally oriented prey to the left and right, respectively, and promote the corresponding turn action when appropriate (by increasing activation on the output neuron assigned the “turn left" action and lowering activation on all other output neurons). When moving neither right nor left, SJ assesses whether it is better to step or jump (by increasing activation on the output neuron assigned the preferred action, and lowering activation on the other). Connections coloured blue varied over different occurrences of the SJ neuron (they do not affect behaviour).