Theory and algorithm of the homotopy BP neural network are proposed in the paper
the error minimization problem in a BP neural network is transfered to the solving system of an associated nonlinear equation
and the globally minimum error and the faster convergence speeds can be obtained by the homotopy BP algorithm.It is checked by results obtained with theoretical analysis and computer simulation.