Does anyone know of work that attempts to build a theorem prover by learning-from-examples? I'm imagining extracting a large corpus of theorems from back issues of mathematical journals, then applying unsupervised structure discovery techniques from machine learning to discover recurring patterns.
Perhaps a model of the "set of theorems that humans tend to produce" would be helpful in proving new theorems.
The unsupervised-structure-discovery bit does seem within the realm of current machine learning.
Yes, and the problem of deciding which lemmas to prove and in general what structure to try for a proof is also far from begin formal. Verifying a proposed proof is the formal bit.
Does anyone know of work that attempts to build a theorem prover by learning-from-examples? I'm imagining extracting a large corpus of theorems from back issues of mathematical journals, then applying unsupervised structure discovery techniques from machine learning to discover recurring patterns.
Perhaps a model of the "set of theorems that humans tend to produce" would be helpful in proving new theorems.
The unsupervised-structure-discovery bit does seem within the realm of current machine learning.
Any references to related work?