The Leeds Institute for Fluid Dynamics (LIFD) has teamed up with the Centre for Environmental Modelling and Computation (CEMAC) to create Jupyter notebook teaching materials for the LIFD CDT.
- Physics Informed Neural Networks
- Image Segmentation
- Autoencoders
- Design Optimisation with Metamodels
PLEASE NOTE YOU MUST CLONE RECURSIVELY (SEE BELOW)
These notebooks are designed to run alongside a teaching module, however the notebook for each topic will include links to further reading where necessary. Each notebook will take about two hours to run through and should run out-of-the-box on home installations of Jupyter notebooks, although they have been primarily designed to work on Leeds-based GPU workstations. These notebooks are designed with automatic checking of Python environment files to remain easy to set up into the future.
As this resource grows, in order not to make the repository unwieldy, this repository is made up of submodules that can be cloned individually.
Some tutorials are so lightweight you can run them on binder. The others we recommend running on your local machine. To get started, either clone this repository (LARGE SIZE) or select a tutorial to clone and run each tutorial separately.
git clone --recursive [email protected]:cemac/LIFD_CDT_ML_NOTEBOOKS.gitthen follow the individual README.md instructions for each notebook.
Instructions for cloning an individual tutorial can be found in the README.md file on that tutorial's GitHub page.
These notebooks can run with the resources provided and the Anaconda environment set up. If you are familiar with Anaconda, Jupyter notebooks and GitHub then simply clone this repository and run it within your Jupyter notebook setup. Otherwise, please read the how to run guide. Individual notebooks have bespoke instructions.
Python
It is recommended you use Anaconda to manage the Python packages required. Some machine-learning libraries are large and if you only wish to run one notebook consider installing the environment provided for that specific notebook. Otherwise, you can install all required packages running the following commands.
conda env create -f <env-file>.yml
conda activate <env-name>
# save yourself some space with one extra command
conda clean -aWhat if I forgot to clone recursively?
Not to worry. In your cloned folder simply run:
git submodule init
git submodule update --init --recursiveHardware
These notebooks are designed to run on a personal computer. Note that some of the techniques demonstrated are computationally intensive, so there may be options to skip steps within certain notebooks depending on the hardware available, e.g. use pre-trained models.
Knowledge
We have assumed some foundational knowledge but links are provided to in-depth information on the fundamentals of each concept.
We hope that this resource can be built upon to provide a wealth of training material for Earth-science machine-learning topics at Leeds.
LIFD_CDT_ML_NOTEBOOKS by cemac is licensed under a Creative Commons Attribution 4.0 International License.
Leeds Institute of Fluid Dynamics, CEMAC, Peter Jimack, Phil Livermore, Jonathan Coney, Donald Cummins, Helen Burns, Andrew Ross, Toni Lassila, Calum Skene, Arash Rabbani