Implicit neural representations for chemical reaction paths

What is it about?

We train neural networks to predict chemical reaction paths and transition states. Compared to the conventional method, this shows (i) resilience to the initial guess, (ii) adaptability to escape local minima, (iii) the ability to capture a complex path on its own, and (iv) potential to generalize to new reactions. It advances the field by offering a flexible alternative to discrete approaches and could unlock building a universal reaction path predictor. The paper should appeal to readers who are broadly interested in machine learning techniques to advance computational chemistry.

Featured Image

Photo by Ivona Rož on Unsplash

Photo by Ivona Rož on Unsplash