Top-down teaching enables task-relevant classification with competitive learning
- Resource Type
- Conference
- Authors
- de Sa, V.; Ballard, D.
- Source
- [Proceedings 1992] IJCNN International Joint Conference on Neural Networks Neural Networks, 1992. IJCNN., International Joint Conference on. 3:364-371 vol.3 1992
- Subject
- Computing and Processing
Components, Circuits, Devices and Systems
Signal Processing and Analysis
Education
Neurons
Unsupervised learning
Supervised learning
Biology
Computer science
Biological system modeling
Hebbian theory
Approximation algorithms
Pattern recognition
- Language
A method of augmenting the basic competitive learning algorithm with a top-down teaching signal which allows task relevant information to guide the development of synaptic connections is described. This teaching signal removes the restriction inherent in unsupervised learning and allows high-level structuring of the representation while maintaining the speed and biological plausibility of a local Hebbian-style learning algorithm. The function of the teaching input is illustrated geometrically, and examples of the use of this algorithm in small problems are presented. This work supports the hypothesis that cortical back-projections are important for the organization of sensory traces during learning.ETX