Skip to content

Latest commit

 

History

History
120 lines (87 loc) · 7.92 KB

README.md

File metadata and controls

120 lines (87 loc) · 7.92 KB

Learning and Processing over Networks

A workshop presented by Michaël Defferrard and Rodrigo Pena at the Applied Machine Learning Days in January 2019.

We suggest you follow the installation guide to setup your own computer. If you don't succeed, you can work in the cloud using binder.

Description

The workshop will introduce the participants to graphs/networks, and provide pointers on how to deal with information defined on such domains.

Participants will be able to see two kinds of graphs: (i) those that are discrete representations of continuous domains, and (ii) those that model relations between entities. Examples of data supported on the networks of the first category are temperature/rain/snow measurements over the Earth, or wind drag over a 3D wing model, or yet category labels (e.g., "land", "buildings", "cars", "people") over a 3D point cloud acquired for autonomous vehicles, land surveys, or indoor mapping. The graphs of the second category include social networks (representing relations between people, and including data about them), the internet (seen through hyperlinks between websites), telecommunication networks (e.g., examining the number of packets per second traveling via the fibers connecting routers), interaction networks (user-product, product-patent, user-user, etc.), road systems, energy networks, and even molecules.

Indeed, networks are ubiquitous in the real world, and, today, data is plenty. In an abstract framework, once the graph and data are defined, the same analysis tools can be used to comprehend that information across various applications. This framework and its tools are the ones we will work with during this workshop. The proposed processing pipeline is versatile and can be fitted to all the examples above, and many more.

The key topics orbiting the theme of this workshop are Network Science, Spectral Graph Theory, Graph Signal Processing, and Machine Learning. After getting acquainted with the basics, participants will choose an application on which to work from a list based on some of the applications mentioned in the previous paragraphs.

Outcome

Participants will learn how to identify network data, how to deal with it, and what can be learned from it. They will know the basics of information processing over networks, and how to devise a machine learning system based on network data. Finally, the hands-on experience will give them the confidence to apply those tools in practice, in applications of their choice.

Prerequisites

  • python programming
  • basic linear algebra
  • no prior knowledge about networks is necessary

Content

This repository contains notebooks for you to practice the presented concepts. They are meant to be worked on while following the slides. The workshop has been prepared as a sequence of presentations and practical sessions.

The repository is made of three branches. The master branch contains the notebooks with instructions and questions for you to follow and answer. The solutions branch adds solutions. The outputs branch contains an executed version of the solutions notebooks. We recommend that you work from the master branch, and consult the solutions from the below links.

  1. Graph and network basics
  2. Network science: basic network properties
  3. Spectral methods: Laplacian eigenmaps and spectral clustering
  4. Spectral representation and filtering

Installation

Binder   Click the binder badge to play with the notebooks from your browser without installing anything.

For a local installation, you will need git, python >= 3.6, jupyter, and packages from the python scientific stack. If you don't know how to install those on your platform, we recommend to install miniconda, a distribution of the conda package and environment manager. Please follow the below instructions to install it and create an environment for the course.

  1. Download the python 3.x installer for Windows, macOS, or Linux from https://conda.io/miniconda.html and install with default settings. Skip this step if you have conda already installed (from miniconda or anaconda). Linux users may prefer to use their package manager.
    • Windows: Double-click on the .exe file.
    • macOS: Run bash Miniconda3-latest-MacOSX-x86_64.sh in your terminal.
    • Linux: Run bash Miniconda3-latest-Linux-x86_64.sh in your terminal.
  2. Open a terminal. Windows: open the Anaconda Prompt from the Start menu.
  3. Install git with conda install git.
  4. Download this repository by running git clone https://github.com/rodrigo-pena/amld2019-graph-workshop.
  5. Create an environment with conda create --name amld2019_graph_workshop.
  6. Activate the environment with conda activate amld2019_graph_workshop (or activate amld2019_graph_workshop, or source activate amld2019_graph_workshop).
  7. Within this environment, install packages by running conda install -c conda-forge python=3.6 jupyterlab geos proj4 libspatialindex, then pip install -r requirements.txt.

Every time you want to work, do the following:

  1. Open a terminal. Windows: open the Anaconda Prompt from the Start menu.
  2. Activate the environment with conda activate amld2019_graph_workshop (or activate amld2019_graph_workshop, or source activate amld2019_graph_workshop).
  3. Start Jupyter with jupyter lab. The command should open a new tab in your web browser.
  4. Edit and run the notebooks from your browser.

Run the test_install.ipynb Jupyter notebook to make sure that the main packages can at least be imported.

If you notice errors of the type 'There is no package called osmnx' or 'There is no package called cartopy', it is because there was an issue while installing them from the requirements.txt file. A possible solution is to install those packages with conda, by running conda install -c conda-forge osmnx cartopy in the terminal, from within the amld2019_graph_workshop environment.

Acknowledgments

The content of the workshop is inspired by the following resources:

License

The content is released under the terms of the MIT License.