Information Technology Reference
In-Depth Information
for which real time computation of large data sets become possible using proper
hardware. The information is determined on connection weights between the layers.
A processing unit consists of a learning rule and an activation function. The
learning rule resolves the actual input of the node by mapping the output of all
direct antecedent and extra external inputs onto a single input value. The activation
function is then applied on the actual input and determines the output of the node.
The output of the processing unit is also described as activation. In the Fig.
2
the
two input nodes are shown in input layer, one output nodes is shown in output
layer. Organizing the nodes in layers resulted in a layered network and the Fig.
2
shows in between input and output layers there are two hidden layers. The inputs to
hidden and hidden to output nodes are connected by weight values that is initialized
during the start of the training and a net input is calculated on which the activation
function is applied to calculate the output. The multilayer perceptron has additional
L
1 hidden layers. The lth hidden layer consists of h(l) hidden units. MLP is
applied to solve wide varieties of interdisciplinary problems like credit scoring
(Khashei et al.
2013
), medical (Pel
≥
á
ez et al.
2014
), food classi
cation (D
ę
bska and
Guzowska-
wider
2011
), forecasting (Valero et al.
2012
), mechanical engineering
(Hwang et al.
2010
), production (Kuo et al.
2010
) etc.
The next section will discuss speci
Ś
cally the various neural network approaches
applied in the development of intrusion detection system.
Fig. 2 ANN architecture
Search WWH ::
Custom Search