someoneserge/theanogeometry

Theano Geometry

The code in this repository is based on the papers Differential geometry and stochastic dynamics with deep learning numerics arXiv:1712.08364 and Computational Anatomy in Theano arXiv:1706.07690.

Who do I talk to?

Please contact Stefan Sommer [email protected]

Installation Instructions

Linux - pip: (recommended)

Install numpy, scipy, theano, jupyter, matplotlib, multiprocess, sklearn:

pip install numpy scipy theano jupyter matplotlib multiprocess sklearn

Use e.g. a Python 3 virtualenv:

virtualenv -p python3 .
source bin/activate
pip install numpy scipy theano jupyter matplotlib multiprocess sklearn

Start jupyter notebook as in

export OMP_NUM_THREADS=1; THEANORC=.theanorc jupyter notebook

Some features, e.g. higher-order landmarks, may require a 'Bleeding-Edge Installation' installation of Theano, see http://deeplearning.net/software/theano/install.html installation instructions.

Linux - vagrant:

Install vagrant and virtualbox, on Ubuntu e.g.:

apt install vagrant virtualbox

In the vagrant folder, run vagrant and ssh to the box:

vagrant up
vagrant ssh -- -L 8888:localhost:8888

Open the url http://localhost:8888/ in your web browser. The notebook password is '12345'

Windows - conda: (recommended)

Install miniconda for Python 3.6 (or higher) from https://conda.io/miniconda.html
Open the now installed 'Anaconda Prompt' program.
Create a new conda environment and activate it by issuing the following commands in the Anaconda prompt:

conda create -n theanogeometry python=3
activate theanogeometry

Use Conda to install the necessary packages:

conda install git numpy scipy theano m2w64-toolchain mkl-service libpython jupyter matplotlib multiprocess scikit-learn

Use git to download Theano Geometry and cd to the directory:

git clone https://bitbucket.org/stefansommer/theanogeometry.git
cd theanogeometry

Start Jupyter:

set THEANORC=.theanorc 
jupyter notebook

Your browser should now open with a list of the Theano Geometry notebooks in the main folder.

Windows - vagrant:

Theano Geometry can be used on Windows with the provided Vagrantfile:

Step-by-step guide for Windows users

Premise:

We are going to use a Virtual Machine, so we need to check that our actual machine allows for it to work. You might want to enter your BIOS and check if virtualization is enabled. Otherwise, just ignore this premise. You can always go back to it if the need arises. An example procedure for enabling virtualization for Lenovo laptops is detailed at: https://support.lenovo.com/dk/en/solutions/ht500006

1. Download and install Vagrant from https://www.vagrantup.com/downloads.html
2. Download and install Virtualbox from https://www.virtualbox.org/wiki/Downloads
3. Download and install a SSH client from http://sshwindows.sourceforge.net/download/
4. Save the file 'Vagrantfile' in the vagrant directory
5. Start a command prompt (e.g. Search for 'cmd')
6. Change directory ('cd') to the directory where you saved the Vagrantfile
7. Run 'vagrant up' and wait a while until the command has finished. If you get an error message stating that "VT-x is disabled", then you need to enter your BIOS and enable virtualization (see premise above).
8. Run 'vagrant ssh -- -L 8888:localhost:8888'.
9. When asked for "passphrase for key", press Enter. When asked for password, enter: 12345
10. Fire up your favourite web browser and open 'http://localhost:8888/'
11. Enter password '12345' and log in.

theanogeometry's Issues

This is a brilliant work

The implementation of geometric structures in this work gives valuable instruction. All the detailed information is included in the examples with specific mathematical structures.

As Theano becomes less popular nowadays, it will benefit other researchers if the Pytorch version of this work is released soon. Thanks for the beautiful job!