Jeffries J W, Conrad M
Department of Computer Science, Wayne State University, Detroit, MI 48202.
Biosystems. 1994;33(3):155-66. doi: 10.1016/0303-2647(94)90001-9.
The structure of the reference neuron model and various fine tuning issues were explored within the domain of face recognition. Several faces were presented for the system to learn. Modified versions of one of the faces were then presented, and the system was asked to identify the faces. Increasing amounts of noise were added to the faces and the system's responses were noted. The system proved very capable of accurately identifying modified faces, even rivaling the ability of the human eye. Many interesting tuning and implementation issues were resolved. The system performed best with minimal numbers of neurons involved in the detection process. Another surprising result is that performance achieved was independent of the choice of features to which the neurons responded.
