Introductory workshop on deep learning "Deep learning on handwritten digits" given at Institut Curie u900 retreat April, 2017. Inspired and adapted from Stanford Vision course cs231n and official TensorFlow tutorial training LeNet on MNIST (https://hub.docker.com/r/tensorflow/tensorflow/).
Docker images (cpu/gpu) available from https://hub.docker.com/r/jcboyd/deep-learning-workshop/.
Install NVIDIA-docker according to https://github.com/NVIDIA/nvidia-docker
docker build -t --build-arg http_proxy=<http_proxy> --build-arg https_proxy=<https_proxy> jcboyd/deep-learning-workshop:[cpu|gpu] .
docker pull jcboyd/deep-learning-workshop:[cpu|gpu]
nvidia-docker run -ti -e http_proxy=<http_proxy> -e https_proxy=<https_proxy> -p 8888:8888 deep-learning-workshop:[cpu|gpu]
Linear model inspired by https://cs231n.github.io/linear-classify/
Notebook arranged for Slideshow with https://github.com/damianavila/RISE
Optionally install Jupyter themes with https://github.com/dunovank/jupyter-themes