with K-means clustering 

This project is an attempt to incorporate unsupervised learning algorithms (K-means clustering) to solve a common architectural problem of clustering non repetitive elements by its geometrical characteristics. As a case study- a 3d non standard tiling was chosen. Vector field is used as a driving force to modify a hexagonal pattern by its magniture and direction. The set of varied tiles is rationalized and partitioned into 5 types by its 1 or 2-dimentional characteristic.

work in progress