Can Machines Intelligently Propose Novel and Reasonable Scientific Hypotheses?

Pengwei Wang, Zhongyuan Wang, Lei Ji, Jun Yan, Lianwen Jin

Abstract: Machine intelligence is attracting increasing attention from both industry and academia. However, the problem of how to make machines innovate novel hypothesis is underexplored. Automatic hypothesis generation can effectively shorten research process. In this work, we try to build an embedding based genetic algorithm to learn "experience" from past data, mine latent semantic information, and then propose the new scientific hypotheses. To our best knowledge, we are the first who propose to use an embedding based genetic algorithm for scientific hypothesis generation. Experiments show that our method outperforms the state of the art.