Information Mandala: Statistical Distance Matrix with Clustering

In machine learning, observation features are measured in a metric space to obtain their distance function for optimization. Given similar that statistically sufficient as population, statistical between two probability distributions can be calculated more precise learning. Provided the observed multi-valued, is still efficient. However, due its scalar output, it cannot applied represent detailed distances feature elements. To resolve this problem, paper extends traditional matrix form, called matrix. The proposed approach performs well object recognition tasks and clearly intuitively represents dissimilarities cat dog images CIFAR dataset, even when directly using image pixels. By hierarchical clustering of matrix, pixels separated into several clusters geometrically arranged around center like Mandala pattern. with Information Mandala.