Gas-type Neural Networks, also known as Kohonen Networks or self-organizing maps, are a type of unsupervised artificial neural network model. These networks are used to classify and visualize complex data in a low-dimensional structure, which makes them useful for data exploration and dimensionality reduction.

The operation of these networks is based on the similarity between input patterns and a set of weight vectors that are adjusted during the training process. These weight vectors represent the nodes of a two-dimensional mesh structure, in which neighboring nodes have similar weights. During training, the input patterns are presented to the network and the weight vectors are adjusted so that similar patterns are clustered at nearby nodes.

Once the network has been trained, it can be used to classify new input patterns into one of the previously defined categories. Gas neural networks are used in various applications, such as data analysis, image segmentation, text classification, anomaly detection, among others.

Gas Neural Networks and SOM (Self-Organizing Maps) networks are terms that refer to the same type of unsupervised artificial neural network.

In fact, the Kohonen Network, which is the original name of these networks, was proposed by the Finnish scientist Teuvo Kohonen in 1982. The term "Self-Organizing Map" was later introduced by Kohonen in 1984, and is a more descriptive term than the original name.

The term "Gas-type Neural Network" is less common than the term "Kohonen Network" or "SOM Network", but it is also used to refer to these networks, particularly in some research studies and in the technical literature.