Re-clustering of Constellations through Machine Learning

Since thousands of years ago, people around the world have been looking up into the sky, trying to find patterns of visible stars’ distribution, and dividing them into different groups called constellations. Originally, constellations are recognized and organized by people’s imaginations based on the shapes of the star distribution. The most two famous groups of stars is the “Big Dipper” and the “Orion”. In modern astronomy, the International Astronomical Union (IAU) has defined constellations as specific areas of the celestial sphere. These areas have their origins in star patterns from which the constellations take their names. In total, there are 88 officially recognized constellations. On the other hand, certain stars are grouped together primarily because they are close to each other and far away from other stars. In other word, one can approximate constellations as the clusters of stars on the celestial sphere. Then it would be quite interesting to see what would constellations (clusters) look like if one uses some totally objective clustering methods regardless of traditions and human imaginations. For example, would the seven stars in the famous constellation ”Big Dipper” still be classified into the same cluster? This gives us an inspiration of re-clustering of constellations using unsupervised machine learning techniques.