Biometric Authentication: A Machine Learning Approach

5.2. Neuron Models

McCulloch and Pitts [237] proposed a binary threshold unit as a computational model for an artificial neuron (see Figure 5.1). Basically, this neuron model can be characterized by the functional descriptions of the connection network and the neuron activation. Each neuron receives input values x_j's, which are propagated through a network of unidirectional connections from other neurons in the network. Associated with each connection, there is a synaptic weight (denoted by w_ij), which dictates the effect of the j-th neuron on the i-th neuron. The inputs to the i-th neuron are accumulated, together with the external threshold θ_i, to yield the net value u_i. The mapping is mathematically described by a basis function. The net value ...