Enhancing the interpretability of the mapper algorithm

The Mapper Algorithm is a powerful tool for representing the topology of a<br/>dataset’s structure as a similarity graph for the purposes of exploratory analysis.<br/>Despite Mapper’s ability to simplify complex high-dimensional data representations,<br/>interpreting the structure of the output graph remains a challenge.<br/>The conventional method of interpreting the Mapper graph by coloring nodes<br/>by features values is infeasible for high-dimensional data due to time limitations<br/>and the potential for subjectivity-related oversights. We present a novel<br/>method to enhance the interpretability of the Mapper algorithm. Specifically,<br/>we propose adapting eXplainable Artificial Intelligence techniques to determine<br/>feature importance, offering both local and global interpretations. Our approach<br/>can be used to assist domain experts in understanding functional differences<br/>across Mapper graphs, enabling them to draw meaningful conclusions from the<br/>graph’s structure.To validate our approach, we conducted experiments on five<br/>real-world medical datasets and the MNIST handwritten digit dataset. Our evaluation<br/>methods consist of a combination of visualization, classification tasks,<br/>and alignment of interpretations to existing literature. The results demonstrate<br/>our method’s effectiveness in providing a means to interpret Mapper graphs by<br/>highlighting the roles of specific features in the graph—such as pixel regions in<br/>MNIST and genes in TCGA datasets.