Thea Ghiselli-Crippa, Paul Munro
A variant of the encoder architecture, where units at the input and out(cid:173) put layers represent nodes on a graph. is applied to the task of mapping locations to sets of neighboring locations. The degree to which the re(cid:173) suIting internal (i.e. hidden unit) representations reflect global proper(cid:173) ties of the environment depends upon several parameters of the learning procedure. Architectural bottlenecks. noise. and incremental learning of landmarks are shown to be important factors in maintaining topograph(cid:173) ic relationships at a global scale.