Artificial Neural Net example in Python using Keras and TensorFlow. This neural net implements a classifier and operates on the 'moons' data set from the scikit-learn package.
To clone this repository:
git clone https://github.com/coledd/keras-example.git
After installing python3, you can install the following dependencies using pip. Alternatively, you may prefer to use your distribution's native package manager for the above packages.
python3 -m pip install --user tensorflow python3 -m pip install --user keras python3 -m pip install --user scikit-learn python3 -m pip install --user matplotlib python3 -m pip install --user numpy python3 -m pip install --user scipy
You may want to do this in the keras-example directory if python2 is your default:
sed -e "1c\/usr/bin/python3" -i ./train.py sed -e "1c\/usr/bin/python3" -i ./evaluate.py
To get a copy of the presentation that was given which included this example:
Here is the "moon data" that the model will train on:
Here is a snapshot of the model loss over the 5000 epoch training:
Here is the resultant decision boundary, with shading that represents the confidence near the boundary: