Office

Via Cairoli 1/4
16124 Genoa – Italy
ph. +39 010 8970500

What are artificial neural networks?

Artificial neural networks (or ANN, artificial neural network) and simulated neural networks (or SNN, simulated neural network) are a subset of machine learning and the core of deep learning algorithms. Their name as well as their structure are inspired by the human brain. Neural networks in fact mimic the way biological neurons communicate with each other.

The nervous system is responsible for everything humans do, every memory and action they perform. Our neurons are optimised to receive information from other neurons, process the information and send the results to other cells. In much the same way as the neuron, the perceptron receives information, synthesises it and sends it through an activation function. This activation function determines whether or not to send the output and to which level. The perceptrons are organised in layers made up of interconnected nodes just like the neurons in the human brain.

Source

 

Types of neural networks

Below are the most common types of neural networks and their use cases:

A- The perceptron is the oldest neural network. This first form of neural network was proposed by American psychologist Frank Rosenblatt in 1958. Although it is now considered outdated, it remains the basis for the design of complex networks.

B- Feedforward networks, or multi-level perceptrons have a layer structure. These levels are :

  • Input layer
  • One or more hidden layers
  • Output layer

In this type of network, no self-coupling or connections with the neurons of its own level are possible. Each neuron therefore has the function of propagating the signal through the network, with a flow of information from the inputs to the outputs.

These models are the basis for computer vision, NLP (natural language processing) and other neural networks.

C- Recurrent neural networks (or RNNs) are identified by their feedback loops. These learning algorithms are mainly used when using time series data to make predictions about future outcomes, such as stock market or sales forecasts.

 

Comparison between neural networks and deep learning

We tend to speak of deep learning and neural networks interchangeably, which can be confusing. Consequently, it is worth noting that the term ‘deep’ in deep learning refers exclusively to the depth of layers in a neural network. A neural network that consists of more than three layers – which would include input and output – can be considered a deep learning algorithm. A neural network that has only two or three layers is only a basic neural network.