Material for running a lab session on atomic neural networks.
The lab in the form of Jupyter notebooks, and is intended for running on Google Colab.
To run this lab you should have a modern browser, it is recommended to run the labs with Chrome. You also need a Google account to use the Colab. If you prefer not to do so, you'll need to install Jupyter notebook, Numpy and TensorFlow 2 on your own machine.
The lab runs on Jupyter notebooks, and you are required to do this lab by writing Python code. You are not required to install anything on your computer, but if you do not know about Python or Jupyter notebooks, it's good that you read something about them beforehand. The lab contains two notebooks to help you to get started with Python and Jupyter notebooks, you can go through them in advance.
In addition, here's a list of resources you can read about
- This tutorial covers Python, Numpy and Matplotlib, which also runs on Colab.
- The official documentation of Python, Numpy, Matplotlib and TensorFlow include both introductory material and detailed information if you'd like to dive deeper
During the lab we will go through the construction and usage of neural networks are a regression tool. You will be guided to build a neural network from scratch. However, it's better if you know something about them in advance.
Here's some material (in addition to the course) that you might be interested to watch:
- 3Blue1Brown's tutorial on neural networks
- A short introduction of neural networks and convolutional neural networks, from computerphile
| Notebook | Link |
|---|---|
| 1. Introduction |  |
| 2. Python basics | |
| 3. TensorFlow basics | |
| 4. Keras models | |
| 5. Atomic neural networks (descriptor based) | |
| 6. Atomic neural networks (graph convolution) | |
| 7. Training ANN with PiNN | |
The lab was originally prepared by Yunqi Shao at Uppsala University as a course lab session for the Computational Quantum Chemistry for Molecules and Materials course in 2019.
The author would like to acknowledge the following people for their assistance:
- Dr. Chao Zhang, for initializing this project and helpful discussions.
- M.Sc. Ageo Meier de Andrade, for his suggestions during the revision of the lab.
- Participants of previous lab session, for their helpful feedback.