Skip to content
Snippets Groups Projects
Commit c34011d7 authored by s1737494's avatar s1737494
Browse files

Update outline.md

parent 496e714c
No related branches found
No related tags found
No related merge requests found
## Force fields in molecular modelling # Force fields in molecular modelling
### Introduction ## Introduction
- Outline to simulations and why we need them ### Outline to simulations and why we need them
- Different methods and timescale - Different methods and timescale
- Pointing out DFT and Classical/Coarse Grain - Pointing out DFT and Classical/Coarse Grain
- Bridging the gap using Machine Learning Interatomic Potentials (MLIPs) - Bridging the gap using Machine Learning Interatomic Potentials (MLIPs)
- Examples of these in the past - Examples of these in the past
### Introduction to MACE ## Introduction to MACE
- What is MACE? ### What is MACE?
- MACE = Message-passing Atomic Cluster Expansion - MACE = Message-passing Atomic Cluster Expansion
- The code that takes in information such as atomic position, energies, forces, stress, etc. and genereates a potential based on that - The code that takes in information such as atomic position, energies, forces, stress, etc. and genereates a potential based on that
- To understand what MACE actually does, we must first visit ACE. - To understand what MACE actually does, we must first visit ACE.
- A condensed theoretical explanation of ACE then how MACE builds on it via NN - A condensed theoretical explanation of ACE then how MACE builds on it via NN
### Trying out MACE ## Trying out MACE
- Introduce MACE Jupyter notebook ### Introduce MACE Jupyter notebook
- Start with running first cell to installe mace and required packages - Start with running first cell to install mace and required packages
- Then make sure to set the cwd to where all the files are - Then make sure to set the cwd to where all the files are
- Going through training parameters and dataset - Going through training parameters and dataset
- Inputs and fileformats - Inputs and fileformats
- Running it once with the default settings and compare with a GAP (Gaussian Approximation Potential ) that was trained on the same data ### Training
-Running it once with the default settings and compare with a GAP (Gaussian Approximation Potential ) that was trained on the same data
### Validating
- Plot against reference parameters (xTB = Semiempirical Tight Binding)
### Varying Inputs
- Change some hyper parameters and training again - Change some hyper parameters and training again
- Importance of hyperparameters - Importance of hyperparameters
- Why changing some won't make a big differences - Why changing some won't make a big differences
- Channels: Does more always more more - Channels: Does more always more more
- To validate:
- Plot against reference parameters (xTB = Semiempirical Tight Binding)
### Using the MACE Model ## Using the MACE Model
- Small introduction into MD (Molecular Dynamics) - Small introduction into MD (Molecular Dynamics)
- Basic theory, ie Newtons 2nd Law - Basic theory, ie Newtons 2nd Law
- Most importanly: Time dependancy and linking back to timescale from introduction - Most importanly: Time dependancy and linking back to timescale from introduction
......
0% Loading or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment