ACCT 002 Lecture Notes - Lecture 11: Constraint Satisfaction, Associators, Positive Feedback
7 Sep 2020
School
Department
Course
Professor
Document Summary
Mcleod: introduction to connectionist modelling of cognitive processes pp. 9- 6: positive feedback ensures that a node that gets ahead is likely to consolidate its advantage, some information is more easily available for retrieval from memory than other information from the same category typicality effect. Problems for distributed representations: the addition of new information doesn"t necessarily cause the loss of old. Evidence for localist information storage: if new information is added, the strength of connections must be changed addition of new information is likely to lead to some loss of old information pp. 51-65. 7: accurate recall of different associations from a pattern association with the hebb rule is possible even when the associations are stored on the same connections. 8: competitive learning is biologically plausible it uses a local learning rule, the information required to change the weight of a connection is available at the axon and the dendrite on either side of the connection.
