AEs-for-ranking-phasefield-feasibility

Autoencoders for ranking the chemical feasibility of novel phasefields, with respect to the dataset on which the models are trained.

Installation Instructions (Windows, Anaconda/Miniconda)

1. Install Anaconda

• If Anaconda or Miniconda is not already installed, it can be installed from the university programs or using the link to installation.

2. Open the Anaconda Prompt

• On windows, search for and open the anaconda prompt window.

3. Create a virtual environment with the required packages

• Type or copy the following line into the command prompt to create a virtual environment:

conda create -n myenv -c anaconda git jupyter pandas numpy scikit-learn matplotlib python=3.9

• 'myenv' can be changed to whichever name you prefer for the environment

4. Activate the environment

• Type or copy the following line into the command prompt to activate the environment:

conda activate myenv

• If you changed the name of myenv in step 3, use this instead of myenv in the above line.

5. Install TensorFlow

• If your workstation has an NVIDIA GPU with CUDA installed, install TensorFlow with the following:

conda install tensorflow

• If you only have CPUs, install TensorFlow with this:

conda install tensorflow-cpu

6. Using GitHub

• Create a GitHub account if needed.
• Ensure you have access to the lrcfmd github repositories

7. Clone the repository

• In the command promt, navigate the directory where you wish to copy this repository. Eg.

cd C:\Users\**YOUR USER**\Documents\

• cd changes directory, dir lists subdirectories within the current directory, mkdir creates a new directory in the directory you are in.
• Clone the repository using the following in the command prompt:

git clone https://github.com/lrcfmd/AEs-for-ranking-phasefield-feasibility.git

• You will then need to login to GitHub to complete the download.

8. Navigate into the working directory

• Change into the working directory with the following line:

cd AEs-for-ranking-phasefield-feasibility/AEs_notebook

9. Unzip the files

• A couple things had to be compressed to save storage.
• Use the following compand to expand the first files:

unzip vectors.zip

• Then change into the DATA directory, and unzip the features:

cd DATA

unzip magpie_tables.zip

cd ..

10. Moving datasets

• If using your own dataset for the training of the autoencoders, move them into the DATA folder.
• This can be done either using the computers GUI and moving the dataset in Files, or by using the following command:

cp C:\Users\**YOUR USER**\path\to\dataset C:\Users\**YOUR USER**\path\to\AEs-for-ranking-phasefield-feasibility\AEs_notebook\DATA

12. Launching programme

• In the AEs_notebook folder, launch jupyter notebook with the following:

jupyter notebook

13. Running the programme

• Follow the instructions within the notebook to run your own ranking.
• It contains an example script, example.ipynb, that uses a pre-assembled dataset to perform a ranking. It may be best to try using this first to ensure it is working, and that the inputs are all understood.