A NEW RECURRENT NEURAL-NETWORK ARCHITECTURE FOR VISUAL-PATTERN RECOGNITION

In this paper, we propose a new type of recurrent neural-network archi tecture, in which each output unit Is connected to itself and is also fully connected to other output units and all hidden units, The propos ed recurrent neural network differs from Jordan's and Elman's recurren t neural networks with respect to function and architecture, because i t has been originally extended from being a mere multilayer feedforwar d neural network, to improve discrimination and generalization powers, We also prove the convergence properties of learning algorithm in the proposed recurrent neural network, and analyze the performance of the proposed recurrent neural network by performing recognition experimen ts with the totally unconstrained handwritten numeric database of Conc ordia University, Montreal, Canada, Experimental results have confirme d that the proposed recurrent neural network improves discrimination a nd generalization powers in the recognition of visual patterns.