Reviews: Premise Selection for Theorem Proving by Deep Graph Embedding

The paper addresses the problem of premise selection for theorem proving. The objective in this line of work is to learn a vector representation for the formulas. This vector representation is then used as the input to a downstream machine learning model performing premise selection. The contributions of the paper are: (1) a representation of formulas as graph; (2) learning vector representations for the nodes and ultimately for the formula graphs; and (3) show empirically that this representation is superior to the state of the art on theorem proving data sets. Pros * The paper is very well written.