Storage of the Distance between Place Cell Firing Fields in the Strength of Plastic Synapses with a Novel Learning Rule

The idea that the hippocampal map has the form of a Euclidean graph of the environment is far from complete. For example, it is plausible that the start of a path is signaled by place cells whose fields are at the rat's current location, but it remains to be shown if the excess activity seen in many place cells while a rat is at a goal location can be used to create a graph-searching algorithm based on realistic neural mechanisms. This chapter focuses on the way in which information acquired during exploration is translated into the static structure of synaptic resistances. It presents a scheme that drastically reduces the bias of producing paths that go through more highly explored parts of the environment. Using a modified version of the Hebbian learning rule, the representation of the environment always tends to become more accurate with increasing exploration time so that eventually the representation asymptotically becomes exact. In addition to its ability to solve an interesting problem, the new learning rule is biologically plausible. It is argued that the novel requirement is the existence of a negative feedback pathway from AMPA receptors to NMDA receptors such that additional strengthening via NMDA receptor activation is inhibited in already strengthened synapses.