Figure 2

Simulations of the adaptive algorithm and measures of learning. a. Knowledge space estimated by the algorithm (top) after 500 simulated trials, shown as five "slices" through the three-dimensional knowledge space cube. Red represents high probability of success, blue high probability of failure, green background = chance level (50%). The estimated knowledge space resembles the actual knowledge surface (bottom) that was used by the simulator module. b. Estimated knowledge volume as a function of the number of trials. Simulated children had a knowledge limit of a rectangular cube of a particular size starting at the origin beyond which they could not progress. It can be seen that the algorithm quickly converges towards the appropriate knowledge volume (approximately the cube root of the imposed limit). c. Here, simulated children had no knowledge limit, but variable learning rates. The program tracked the progressive increase in the knowledge volume. (Knowledge started at 0.5 on each dimension).