Automatic data‐driven stoichiometry identification and kinetic modeling framework for homogeneous organic reactions

Abstract

Data-driven and knowledge-driven methods are two approaches used in studying reaction kinetics. This article proposes a hybrid-modeling framework for homogeneous synthesis reactions, which combines the advantages of high level of automation in the data-driven approach and improved accuracy in the knowledge-driven approach. A constrained enumeration method is proposed to generate possible candidate stoichiometries, and dynamic response surface methodology, target factor analysis, and mass balance are used together for identifying stoichiometries one-byone, without the necessity of an expert-generated candidate list. Then, the previously screened stoichiometries are formed into different groups that represent candidate reaction systems, and the group (or groups) with the greatest likelihood will be identified, based on kinetic fitting and reaction dynamic criteria. This framework has been demonstrated by several examples of different reaction systems. The true reaction stoichiometries are all correctly identified, and the accurate kinetic models are obtained, showing satisfactory performance of the proposed method.

Generalizing property prediction of ionic liquids from limited labeled data: a one-stop framework empowered by transfer learning

“…These DNN-based methods have garnered signicant attention due to their ability to overcome the limitations of conventional models and achieve high accuracy in predicting complex tasks. [28][29][30][31] The growth of deep learning (DL) has offered excellent exibility and performance to learn molecular representations from data, without explicit guides from experts. [32][33][34] Typically, a sufficiently large labeled training dataset is desirable for developing DL approaches.…”

A Transformer-Convolutional Neural Network Based Framework for Predicting Ionic Liquid Properties

“…DNN-based ML systems have aroused great interest by overcoming obstacles of conventional models and obtaining high prediction quality for complex tasks. [35][36][37][38] The growth of deep learning (DL) has provided excellent flexibility and performance to learn molecular fingerprints from data, without explicit guides from experts. [39][40][41] In our previous work, a DNN based recommender system (RS) for predicting the solute-in-IL infinite dilution activity coefficient (γ ∞ ) was developed without any manually designed fingerprint.…”