Artificial Neural Networks in a Nutshell

An ANN is a simplified model of the human brain consisting of several layers of neurons that pass signals to each other depending on the input signals that they 

FIGURE 4.10 Schematic image of an artificial neuron. The input data x are calculated with their connective weight w to form the Net value of the neuron. A transfer function is applied to mimic the threshold of the biological neuron. The out value represents the outcome of the process, which is fed to another artificial neuron. 

The Net value undergoes a transfer function that mimics the signal threshold from the biological neuron. A typically used transfer function has a sigmoidal shape expressed by 

We can introduce this function to change the bias of the outgoing signal outj of a neuron. 

With introducing the factor a for the Net values of the neuron we are able to define the steepness of the function. The 9 value affects the relative shift of the function on the x-axis. The out value may act as an input value for another neuron or as final resnlf. An ANN is created by connecting multiple neurons in multiple layers. Typically, each neuron in the inpnt layer receives the input data and is connected to each nenron of the next layer this architectnre repeats down the output layer, which finally confains fhe onfpnf valnes (Fignre 4.11). The layers between the input and ontpnt layer are referred to as hidden layers, and the nnmber of hidden layers as well as the number of neurons in each layer can be adapted to the task. 

---