Follow Us On

Machine Learning Force Fields Unlock Atomic Simulations

Machine Learning Force Fields Unlock Atomic Simulations

Speaker
Abstract

This talk delves into the emerging role of machine learning (ML) force fields in materials science and biology, exploring their potential to revolutionize our understanding of complex atomic and molecular interactions. We will discuss the limitations of classical force fields and how ML techniques, such as deep learning and kernel methods, can model potential energy surfaces more accurately and efficiently. By learning from quantum mechanical calculations, ML force fields can provide insights into chemical reactions, protein folding, and material properties, guiding experimental research and accelerating drug discovery. The presentation will emphasize the transformative potential of ML force fields and the importance of continued interdisciplinary research to harness their full capabilities.

Outline:
  1. Introduction
    • Molecular dynamics
    • Parametric force fields
    • Ab initio molecular dynamics
  2. Representing Chemical Environments
    • Local atomic descriptors
    • Linear force fields
    • Gaussian Process force fields
  3. Neural network force fields
    • Multi Layer Perceptron Force Fields
    • Graph neural networks
    • Eqivariant graph neural networks
    • Transformer architectures
  4. Applications showcase
Schedule