Hand-Written Digits Recognition by Graph matching and Annealing Neural Networks

Mean field annealing(MFA) is a promising tool in optimization and a neural network model based on the graph matching have attracted attention due to a number of benefits over conventional recognition models. We present a neural network model for hand-written digits recognition using graph matching and two annealing techniques, MFA and one-variable stochastic simulated annealing(OSSA). OSSA makes it possible to evaluate the equilibrium spin average value effectively by Monte Carlo technique. In this paper hand-written digits recognition can be formulated as elastic graph matching, which is performed here by annealing techniques of matching cost function. Our model provides not only the function of recognition but also the segmentation ability such that input characters are correctly recognized and segmented simultaneously even if they are touching, connected, and defected by noise. Some simulation results show the capability of our model and the characteristics of MFA and OSSA. Key-Words: Pattern recognition, Segmentation, Optimization, Mean field annealing, Annealing neural networks