Hippocampal Anatomy Supports the Use of Context in Object Recognition: A Computational Model
Figure 6
Error for each of the four output layers of the Baseline network on various sample tasks, as a function of crossconnection weighting. Crossconnection input refers to LEC input to anterior hippocampus and MEC input to posterior hippocampus. Higher relative weight multiplier values mean stronger MEC input to posterior and stronger LEC input to anterior streams. (a) Object output error on noisy and partial cue tests (where both object and context are noisy or partial, resp.) as a function of crossconnection strength. (b) OBCG output error on noisy and partial cue tests (here the noise and partial are only in the context) as a function of crossconnection strength. (c) Same as A, except the error is measured at the context output layer. (d) Same as B, except only the object is noisy or partial, and the error is measured at the CBOG output layer. Error bars are standard errors of the mean.
