Cognitive Mapping
Main Goals
He claims to present a real-time, biologically plausible neural network approach to purposive behavior and cognitive mapping.
Architecture
The system is composed of (a) an action system, consisting of a goal-seeking neural mechanism controlled by a motivational system; and (b) a cognitive system, involving a neural cognitive map.
Places refer to discernable regions in the maze. Views refer to the sights of places as seen from another place. The cognitive map associates Place representations with View representations.
Building the Cognitive Map
Place trace Xi may become associated with the views of other places to form long-term associations Vij between Place i and View j.
The cognitive map is updated whenever a COMPARATOR detects a mismatch between predicted and actual inputs.
Reading the Cognitive Map
Recurrent reinjection the signal representing View j as predicted by Place i according to Vij, into the representation of Place j. Place j now predicts View k according to Vjk and the signal representing View k is reinjected into the representation of Place k. The process continues until the representation of the goal is eventually activated.
Fast time PREDICTIONS are produced in advance of what occurs in real time, when the information stored in the cognitive map is used to predict the remote future.
The active scanning of alternative pathways at choice points, reflects the animal's generation and comparison of different expectancies.
Achievements
Computer simulations show that the network successfully describes LATENT learning and DETOUR behavior in mazes.
The cognitive map allows the combination of information about spatially adjacent places, and thereby the inference of the connections to remote places (by reinjection).
In addition the network can be applied to problem-solving (means-end field) paradigms such as the Tower of Hanoi puzzle (navigate the transition tree).
Limitations
Description is too phenomenological with not enough biological grounding.
Prepared by Alex Guazzelli and Fernando Corbacho (aguazzel/corbacho@pollux.usc.edu)