Neural Networks are generated from the Spectra Classifier module. Similarly as in Spectra Projector, Neural Network cannot be directly opened from the program desktop. Once the network has been generated, classification results cannot be altered. If you want to remove a spectrum (symbol) from a network, you must remove this spectrum from the input data, and then launch a new classification process. An unlimited number of Neural Networks windows can be opened at any given time in the program.

A Neural Networks window displays a topological map of neurons that are drawn as rectangle. Every neuron has a minimum of two and maximum of six neighboring neurons. Spectra are displayed as symbols or numbers within neurons.

Although neurons are displayed in a regular 2D lattice, the actual Euclidian distances between neurons vary. To show the approximate distances between neighboring neurons, the borders of the neuron are shown using different thickness. The line thickness is proportional to the distance between immediate neighbors. The thinner the line, the closer together are the neurons.

Mass Frontier allows you to save and open neural networks that have been generated with particular initial settings (transformation method and lattice dimension).

It is an important part of classification analysis to know which spectrum is represented by a symbol on a lattice. As Mass Frontier links corresponding modules, spectra with structures or chromatographic components can be recalled from any neural network.