Introduction

https://fgnt.github.io/2019_ad_xidian/

Theoretical exercise

https://github.com/fgnt/2019_ad_xidian/blob/master/theory/exercise.pdf

Practical exercise

Start a jupyter notebook server in the poolroom:

source /upb/scratch/users/c/cbj/py37/bin/activate
cd ~/ && jupyter notebook

For more details see: https://fgnt.github.io/python_crashkurs_doc/include/poolroom.html

Small numpy introduction: https://fgnt.github.io/python_crashkurs_doc/include/numpy.html Numpy cheat sheet: https://git.cs.upb.de/chthiel/python-tutorial/blob/master/cheat_sheets/Numpy_Python_Cheat_Sheet.pdf

Download the exercise:

Download the git repository

git clone https://github.com/fgnt/2019_ad_xidian.git

Now you can find in your home directory a notebook to start the exercise (~/2019_ad_xidian/practice/ad_template.ipynb) and a python script that contains some helper functions (~/2019_ad_xidian/practice/ad_helper.py).

Alternative: Open https://raw.githubusercontent.com/fgnt/2019_ad_xidian/master/practice/ad_template.ipynb in a browser and safe the file.

Final task

Extend the code in jupyter notebook from the practical exercise to a full neuronal network (NN) framework and train a NN on the MNIST data. Can you reach 98% accuracy?