The authors target the problem of curve fitting on data samples for the purpose of subsequent interpolation. This is what backpropagation was developed for but its dependency on initial conditions and net topology aff...
详细信息
The authors target the problem of curve fitting on data samples for the purpose of subsequent interpolation. This is what backpropagation was developed for but its dependency on initial conditions and net topology affects its robustness. Here the authors present a different method which is based on an analysis of possible properties of the internal representations developed as a result of learning. Thus they introduce some additional constraints concerning direct control on internal representations. This method incorporates properties of supervised as well as unsupervised learning in the fitting problem.
A two-phase backpropagation algorithm is presented. In the first phase the directions of the weight vectors of the first hidden layer are constrained to remain in directions suitably chosen by pattern recognition, dat...
详细信息
A two-phase backpropagation algorithm is presented. In the first phase the directions of the weight vectors of the first hidden layer are constrained to remain in directions suitably chosen by pattern recognition, data compression, or speech and image processing techniques. Then, the constraints are removed and the standard backpropagation algorithm takes over to further minimize the error function. The first phase swiftly situates the weight vectors in a good position which can serve as the initialization of the standard backpropagation algorithm. The generality of its application, its simplicity, and the shorter training time it requires, makes this approach attractive.< >
暂无评论