Demonstration of machine learning models trained on the Fashion-MNIST data set. The first part of the notebook visualises clusters through a dimensionality reduction using UMAP.
The notebook will run in a Conda environment, as described below.
At the terminal prompt, create a new conda environment:
conda create -n julita python=3.6 numpy scipy scikit-learn numba jupyter matplotlib seaborn tensorflow plotly pandas
Activate that environment:
source activate julita
The following additional packages need installing using pip, after activating the environment in the previous step:
pip install xgboost
pip install umap-learn
Running the notebook requires both the Fashion_MNIST data and the Fashion_MNIST demonstration image.
The original Fashion_MNIST data can be obtained from the following repository:
https://github.com/zalandoresearch/fashion-mnist
The Fashion_MNIST data will then be in the subdirectory
data/fashion/
as four files *.gz, and the demonstration image will be:
doc/img/fashion-mnist-sprite.png
Please go to the Machine Learning Demonstation file to run the script and compare models.
This project is licensed under the MIT License - see the LICENSE.md file for details