Using sequence to sequence neural networks for solving similar mathematical problems
Ali Davody, Mihai Sebastian Baba
Deep neural networks have enjoyed great success in recognizing patterns among large datasets. On the
other hand, proofs of lots of mathematical theorems are very similar to each other. In this paper, by representing
problems as directed graphs, we provide a concrete definition of similarity notion between problems. Then we examine the performance of deep sequential models to predicting solutions of similar mathematical problems.