Information Lattice Transform

The information lattice transform (ILT) converts raw data into a learned concept space in a domain-agnostic way by drawing on information lattice learning, a directly human-interpretable approach to machine learning with group-theoretic and information-theoretic foundations. The knowledge discovered by constructing the learned concpet space is already scientifically valuable. Moreover, once in the ILT domain, one can perform several downstream tasks including classification, compression, and co-creativity. State-of-the-art performance for classification, information-theoretically optimal performance for compression, and compelling performance for creativity have previously been demonstrated. We further argue that the ILT domain is suitable for continua problems for simulation through solutions of differential equations, and for calibration for digital twins. This can have implications for scientific and engineering fields including particle physics, fluid dynamics, materials discovery, and semiconductor manufacturing.