A typogenetic application of random seeded crystal learning
Stuart H. Rubin · 2003
A next-generation expert system for use in the rational design of pharmaceuticals is described. An example making use of a typogenetic abstraction of the problem is provided. He addresses the problem of predicting molecular keys from an enzyme's primary structure, i.e., amino acid sequence. Arbitrary sequences of bases are called strands. Enzymes, consisting of sequences of amino acids, perform domain-specific functions upon the strands. Unlike real genetics, an individual amino acid's contribution to the enzyme's overall function is considered to be context-free for the sake of simplicity. The typogenetics consists of fifteen types of amino acids. Random seeded crystal learning methods are discussed. All acquired knowledge is represented in the form of situation-action rules. Here, random seeds of knowledge aggregate into regular crystalline geometries under transformational operators acting over the course of time. Next-generation expert systems have the capability for learning by transformational analogy. These systems have a capability for learning through the transformation of knowledge across similar domains implying better explanation facilities and a reduced requirement for case memory. They are also built with operating system principles for the efficient organization of memory.>