README.md

Molecular Design of Solvents Coupling Variational Autoencoder (VAE) and Nonlinear Programming (NLP)

This repository contains the data and scripts involved in the publication:

A Novel Machine Learning-Based Optimization Approach for the Molecular Design of Solvents

NOTE: Part of the scripts used for VAE model development was adapted from the Chemical VAE GitHub repository.

Requirements

Library (Python 3.5)

In addition to the libraries used in Chemical VAE, other required libraries include:

• Pyomo: nonlinear programming
• scikit-learn: data processing
• matplotlib & seaborn: visualization

Software

• COSMOtherm: thermodynamic prediction

Authors

• Zihao Wang
• Prof. Dr.-Ing. Teng Zhou
• Prof. Dr.-Ing. Kai Sundmacher