A new organization of very large database

A theory of a new type of database, an attractor database, is proposed which features compressing a huge base relation in a relational database into a tiny one that catches the characteristics of the original relation efficiently. This type of database is intended to simulate the way human brains store data. This type of database is especially useful in a replicated, widely distributed, very large database where communication costs are high. The development of this scheme is based on a theory of class 2 dynamical systems, defined as single point attractor dynamical systems. An approach using ω-orbit finite automata is a special case of this method. This scheme encodes relations in a database. The encoding procedure stores a relation or a tuple (if the tuple is large) in an attractor of a class 2 dynamical system. A query is implemented in two stages: retrieve an attractor and then query. Several retrieval algorithms and inference algorithms of a class 2 dynamical system from a given relation are introduced