The Target Switching Algorithm, The Offset Algorithm, and The Sequential Learning Algorithm - Constructive Algorithms for Multilayer Perceptrons

Short Description

The Target Switching, Offset, and Sequential Learning Algorithms are another constructive algorithms for multilayer perceptrons like algorithms we familiar with such as Upstart, Perceptron-Cascade, Tiling, Tower, and Pyramid. They offer procedures for constructing and training a feedforward neural networks. The rationale for each algorithm with detailed explanation of algorithms, comparisons with other constructive algorithms, and possible improvements are to be discussed.

Reference

The Target Switch Algorithm: A Constructive Learning Procedure for Feed-Forward Neural Networks. Campbell, C. and Vicente C. P. in "Neural Computation". Vol. 7, pp. 1245-1264. 1995.
The Offset Algorithm: Building and Learning Method for Multilayer Neural Networks. Martinez, D. and Esteve, D. in Europhysics Letters, Vol. 18, No. 2, pp. 95-100. 1992.
A Convergence Theorem for Sequential Learning in Two-Layer Perceptrons. Marchand, M., Golea, M. and Rujan, P. in Europhysics Letters, Vol. 11, No. 6, pp. 487-492. 1990.