Generative reasoning for symbolic discovery

An apparatus comprising:

• a memory;
• a non-transitory computer readable medium comprising computer executable instructions;
• and
• at least one processor, coupled to said memory and said non-transitory computer readable medium, and operative to execute said instructions to be operative to:
• instantiate a computerized generative reasoner, a computerized model inference engine, and a computerized prediction module;
• provide a background theory applicable to a scientific problem as input to said computerized generative reasoner;
• with said computerized generative reasoner, produce a plurality of provable conjectures applicable to said scientific problem, based on said input to said computerized generative reasoner;
• provide said plurality of provable conjectures and a set of input training data to said computerized model inference engine;
• with said computerized model inference engine, fit said input training data to said plurality of provable conjectures to obtain at least one candidate symbolic model reflecting scientific laws associated with said scientific problem;
• reduce a search space of a computerized prediction module by providing to said computerized prediction module said at least one candidate symbolic model, wherein the reducing of the search space uses a combination of:
• constraints, a grammar to restrict the search space to grammatically well-formed formulas, and a restriction of the search space to only derivable formulas;
• provide to said computerized prediction module new data;
• with said computerized prediction module, search in said reduced search space to make a prediction related to said scientific problem based on said new data and said at least one candidate symbolic model;
• and
• mitigate a physical system subject to a physical problem based on said at least one candidate symbolic model, wherein said scientific problem comprises said physical problem.

A non-transitory computer readable medium comprising computer executable instructions which when executed by a computer cause the computer to perform a method of:

• instantiating a computerized generative reasoner, a computerized model inference engine, and a computerized prediction module;
• providing a background theory applicable to a scientific problem as input to said computerized generative reasoner;
• with said computerized generative reasoner, producing a plurality of provable conjectures applicable to said scientific problem, based on said input to said computerized generative reasoner;
• providing said plurality of provable conjectures and a set of input training data to said computerized model inference engine;
• with said computerized model inference engine, fitting said input training data to said plurality of provable conjectures to obtain at least one candidate symbolic model reflecting scientific laws associated with said scientific problem;
• reducing a search space of said computerized prediction module by providing to said computerized prediction module said at least one candidate symbolic model, wherein the reducing of the search space uses a combination of:
• constraints, a grammar to restrict the search space to grammatically well-formed formulas, and a restriction of the search space to only derivable formulas;
• providing to said computerized prediction module new data;
• with said computerized prediction module, searching in said reduced search space to make a prediction related to said scientific problem based on said new data and said at least one candidate symbolic model;
• and
• mitigating a physical system subject to a physical problem based on said at least one candidate symbolic model, wherein said scientific problem comprises said physical problem.