Figure 3.

Development and quality of firing fields as produced by our hierarchical SFA network. Each row depicts the activity of three cells after exploration of a square environment (white cue card at the north wall) for 30 s and 1, 2, 4, and 8 min of simulation time. Since SFA is not an online algorithm that can be stopped and continued, five individual networks were trained anew along incrementally longer sections of the same overall trajectory. The cells depicted along each column are thus not the same output signals, but rather hand-matched signals produced by the five different SFA networks. This matching was done in order to help compare the quality of place fields over time, and it can be seen that stable and distinct fields develop within 2 min of simulation time.