A Study on Interactive Proteomic Data Clustering: A Comparison between Self-Organizing Map and Neural Gas

Abstract

The neural network algorithms Self-Organizing Map (SOM) and Neural Gas (NG) use unsupervised competitive learning. These methods have the critical virtue of preserving the topological structure of the data, which means that data that are close in the input distribution are mapped to neighboring positions in the network or output. This characteristic makes them intriguing to investigate in terms of data clustering. A crucial characteristic analyzing vast amounts of data manually can be challenging and time-consuming. As a result, technologies for analyzing and visualizing massive multidimensional data sets are required.  We introduce a method for comparing and visualizing the SOM and NG in this chapter. We describe these algorithms first, and then we make a pictorial comparison between them. The protein mass spectrometry data clustering is then interpreted using these visualization approaches.