Neurons, Dendrites, and Pattern Classification Gerhard X. Ritter 1 , Laurentiu Iancu 1 , and Gonzalo Urcid 2 1 CISE Dept., University of Florida, Gainesville, FL 32611-6120, USA {ritter, liancu}@cise.ufl.edu 2 Optics Dept., INAOE, Tonantzintla, Pue. 72000, Mexico gurcid@inaoep.mx Abstract. Computation in a neuron of a traditional neural network is accomplished by summing the products of neural values and connection weights of all the neurons in the network connected to it. The new state of the neuron is then obtained by an activation function which sets the statetoeitherzeroorone,dependingonthecomputedvalue.Weprovide analternativewayofcomputationinanartificialneuronbasedonlattice algebra and dendritic computation. The neurons of the proposed model bear a close resemblance to the morphology of biological neurons and mimic some of their behavior. The computational and pattern recog- nition capabilities of this model are explored by means of illustrative examples and detailed discussion. 1 Introduction Various artificial neural networks (ANNs) that are currently in vogue, such as radial basis function neural networks and support vector machines, have very little in common with actual biological neural networks. A major aim of this paper is to introduce a model of an artificial neuron that bears a closer re- semblance to neurons of the cerebral cortex than those found in the current literature. We will show that this model has greater computational capability and pattern discrimination power than single neurons found in current ANNs. Since our model mimics various biological processes, it will be useful to provide a brief background of the morphology of a biological neuron. A typical neuron of the mammalian brain has two processes called, respec- tively, dendrites and axons. The axon is the principal fiber that forms toward its ends a multitude of branches, called the axonal tree. The tips of these branches, called nerve terminals or synaptic knobs, make contact with the dendritic struc- tures of other neurons. These sites of contact are called synaptic sites. The synaptic sites of dendrites are the places where synapses take place. Dendrites have many branches that create large and complicated trees and the number of synapses on a single neuron of the cortex typically ranges between 500 and 200,000. Figure 1 provides a simplified sketch of the processes of a biological neuron. It is also well-known that there exist two types of synapses; excitatory synapses that play a role in exciting the postsynaptic cell to fire impulses, and A. Sanfeliu and J. Ruiz-Shulcloper (Eds.): CIARP 2003, LNCS 2905, pp. 1-16, 2003.  Springer-Verlag Berlin Heidelberg 2003