Further details provided in the documentation within the final_model.py file.
Project Novel is an attempt to use a recurrent neural network (with LSTM cells) to learn the patterns in, and concequentially generate, molecules in the form of SMILES strings.
Here's a quick article I wrote on Medium about the steps I took to build this somewhat dingy project 😄 https://towardsdatascience.com/lets-make-some-molecules-with-machine-learning-%EF%B8%8F-429b8838e3ef
The ultimate goal of this project was to train a generative recurrent neural network to create new molecules, in likeness of those presented in the given dataset. The validity of these generated molecules are questionable and left ambiguous, as future projects will not only address this opportunity, but also that of specified and/or desired properties. These properties can include both material, structural, chemical, themodynamical, etc., including conductivity, strength, transparency, heat-resiliency, etc. This can have potentially earthshattering consequences on the very way we go about scientific and technological progress, saving invaluable amounts of money and significant amounts of time. Upwards and onwards, always and only! 🚀