d241: Deep learning in chemical design using a data-driven continuous representation of molecules

Deep learning in chemical design, using a continuous representation of molecules.

A method to convert discrete representations of molecules to and from a multidimensional continuous representation. This generative model allows efficient search and optimization through open-ended spaces of chemical compounds. Deep neural networks trained on hundreds of thousands of existing chemical structures to construct two coupled functions: an encoder and a decoder. The encoder converts the discrete representation of a molecule into a real-valued continuous vector, and the decoder converts these continuous vectors back to the discrete representation from this latent space

Paper: https://arxiv.org/abs/1610.02415 (7 Oct 2016)

PDF: https://arxiv.org/pdf/1610.02415.pdf