Dresden 2020 – wissenschaftliches Programm

Die DPG-Frühjahrstagung in Dresden musste abgesagt werden! Lesen Sie mehr ...

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

AKPIK: Arbeitskreis Physik, moderne Informationstechnologie und Künstliche Intelligenz

AKPIK 1: AKPIK Talks

AKPIK 1.4: Vortrag

Montag, 16. März 2020, 17:30–17:45, HSZ 301

Artificial Intelligence for the discovery of new molecules — •James Nelson and Stefano Sanvito — Trinity College Dublin, Dublin, Ireland

Many problems facing society today have their solution partially in the discovery of new molecules and materials, i.e. designing energy-efficient technology and drug discovery. However, the space of possible materials is enormous and the known set of compounds only scratches the surface. We propose to use Generative Adversarial Networks (GANs) to help explore this chemical space and suggest new potential compounds. Here we have two neural networks, with one network suggesting potential molecules and another acting as a critic. The two networks compete against each other, resulting in the generative network being able to produce realistic molecules.

In this talk, I'll show our preliminary results in applying GANs to molecules. In particular, we have: designed a unique GAN that incorporates structural information, been able to reproduce thermal distributions of a diverse set of molecules and found chemically sound molecules and that maximise some criteria. This work paves the way towards the use of Artifical Intelligence as a tool to search through chemical space.