Abstract: A method of deriving a correlation between observed values from e.g. a process, where initial values are derived from ingredients or process parameters and a second value from a resulting product. A number of candidate mathematical functions are generated in a grid-type flow structure in which tokens flow through nodes, each node determining a mathematical function, so that a token defines a number of mathematical functions and an order thereof. Tokens may be combined so that the functions include multiple variables. A best, resulting mathematical function may be determined and the grid structure amended to favour this function and the method may be repeated.

Abstract: A method of deriving a correlation between observed values from e.g. a process, where initial values are derived from ingredients or process parameters and a second value from a resulting product. A number of candidate mathematical functions are generated in a grid-type flow structure in which tokens flow through nodes, each node determining a mathematical function, so that a token defines a number of mathematical functions and an order thereof. Tokens may be combined so that the functions include multiple variables. A best, resulting mathematical function may be determined and the grid structure amended to favour this function and the method may be repeated.