Discovering User Interests by Document Classification - Mining and Analyzing Social Networks - page 145

Fig. 6 Neural network

Each connection between unit i and unit j is defined by the weight w ij
determining the effect which the signal of unit i on unit j . Suppose input unit,
hidden unit and output unit is denoted as x, h, y respectively. In the topology, unit
y is the composition of other units h which in turn are the compositions of others
units x . The composition of a unit is represented as a weighted sum which will be
evaluated to determine the output of this unit. If such unit is the output unit, its
output is the output of neural network. The process of computing the output of a
unit includes two following steps:

− An adder so-called summing function sums up all the inputs multiplied by

their respective weights. It is essential to compute the weighted sum. This
activity is referred to as linear combination.

− An activation function controls the amplitude of the output of the neuron.

This activity aims to determine the output.

x 0

θ k (bias)

w 0k

x 1

w 1k

output

s k

Σ μ(.)

y k

x 2 w 2k

Summing
function

Activation
function

w ik

x i

input

Fig. 7 The process of computing the output of a unit