Question

4. Compare the radial basis functions neural network and back propagation in term of various aspects related to training, convergence, and applications.

Answer 1

Aspect	Radial Basis Functions RBFNN	Back propagation Network BPNN
Training	RBF is a 3 layer feed forward network with only one single layer between input and output layer. In training vectors are propagated to hidden layer where where RBF function. Turn input to output adjusting weight of input to hidden layer .	Here input is propagated forward through the Network to get output layer response . Then sensitivity is propagated backward to reduce error . During this weight of all hidden layers is modified and with iterations output result is improved.
Training Time	Have shorter training time as it skips feedback process that occurs in Back propagation network	Being multilayer network and performing back propagation it have drawback of taking longer training times
Convergence	Have faster convergence rate than BPNN	Have slower convergence rate than RBFNN
Applications	Used for Approximation Modelling , weather forecasting	used when underlying feature of data is found inside very large size data sets like image recognition in Imaging processing systems