Learning by dilution in a Neural Network
arXiv:cond-mat/9611130 · doi:10.1088/0305-4470/30/22/014
Abstract
A perceptron with N random weights can store of the order of N patterns by removing a fraction of the weights without changing their strengths. The critical storage capacity as a function of the concentration of the remaining bonds for random outputs and for outputs given by a teacher perceptron is calculated. A simple Hebb-like dilution algorithm is presented which in the teacher case reaches the optimal generalization ability.
LaTeX, 15 pages incl. figures