A comprehensive tutorial on the SOM-RPM toolbox for MATLAB

We present the SOM-RPM Toolbox for MATLAB, which is an interactive command line implementation of the self-organizing map with relational perspective mapping (SOM-RPM) algorithm. SOM-RPM has shown considerable utility for the interpretation of complex hyperspectral data. In essence, it provides a means for interactively exploring similarities between pixels (based on their spectral information) through the so-called similarity map. This manuscript provides an overview of the theoretical underpinnings of SOM-RPM, followed by a detailed description of the SOM-RPM toolbox structure. We supplement these sections with a demonstrative case study using time-of-flight secondary ion mass spectrometry (ToF-SIMS) imaging data as the subject of the analysis. This case study emphasizes the interactive nature of the toolbox and the method itself, which allow for exploration of the data based on the SOM-RPM model. It also highlights the analytical potential of the approach. Our primary aim is to make the SOM-RPM method more accessible to the broader scientific community. This manuscript provides sufficient content for a non-expert in machine learning to be able to utilize SOM-RPM for exploratory analysis of their hyperspectral data. The toolbox, and associated documentation, is available through the linked data repository. • SOM-RPM is an evolved implementation of the self organizing map (SOM). • SOM-RPM is an unsupervised machine learning method developed for ToF-SIMS data. • The MATLAB toolbox is demonstrated using hyperspectral data from inkjet prints. • SOM-RPM is a tool for exploratory analysis of many types of hyperspectral data.