Figure 4. Benefits of Selective Sampling.

Note. In the active-sampling model, the number of features sampled (A) negatively covaried with the slope of the learning curve (B). This reflects the models ability to efficiently explore unsampled features (i.e., by considering the number of times each feature has been observed; Equation 11), and capitalize on the single reliable feature in the simulation environment. The selective sampling model (B) learned more quickly than a comparable model in which all stimulus features were always sampled (C).