Beginning Anomaly Detection Using

 Artificial Neural Networks

Artificial neural networks are layers of interconnected nodes, or artificial neurons, that 

function in a way inspired by biological neural networks. Figure 

3-1

 shows an example of 

a neuron.

Figure 3-1.  An example of what a neuron can look like

Chapter 3   IntroduCtIon to deep LearnIng

---

75

Inputs are taken in through the dendrites, after which the neuron decides whether 

or not to fire. Upon firing, the neuron sends a signal down the axon to its terminal axons, 

where the signals are output to any other neurons. This transfer of signals is called a 

synapse, which is modeled in Figure 

3-2

.

We use a similar concept in artificial neural networks (Figure 

3-3

).

Figure 3-2.  How two neurons might connect to form a chain and transfer signals 

through that connection. The terminal axon of the first neuron connects to the 

dendrites of the second neuron

Figure 3-3.  How an artificial neuron in an artificial neural network can function. 

This mimicry of the biological neuron is the basis of artificial neural networks

Chapter 3   IntroduCtIon to deep LearnIng

---

76

In the case of this artificial neuron, we find the dot product between the input 

vector