Wroclaw University of Science and Technology Faculty of Computer Science and Management

medical.ml Students Research Group

Medical.​ml workshops on data processing tools:

• pandas - processing tabular data
• dask - parallel and cluster-distributed processing

At some point in the future...

You can set up all requirements in your own environment without building Docker images. You'll need:

1. Conda or another python distribution with packages:
   • pandas 0.20.0+
   • numpy 1.13.0+
   • dask 0.16.0+ (preinstalled with conda, make sure you have the latest version)
   • bokeh 0.12.10+ (preinstalled with conda, make sure you have the latest version)
   • scikit-learn
2. (Optional) Graphviz:
   • Download from https://www.graphviz.org/download/ (or apt-get install graphviz on Linux)
   • Install python package graphviz.

     Note 1: install with pip install graphviz rather than with conda install. Conda sometimes messes this up. Note 2: On Windows you may need to manually set up your environment path, or do so right in your code with:

      import os
      os.environ["PATH"] += os.pathsep + 'disk/install/path/Graphviz2.38/bin/' 

     Note 3: The Graphviz provides a handy functionality to the Dask library but is not at all required for it to run properly, so it's up to you to decide whether to install it or not.

Workshop content is located in the ./notebooks/ directory in a form of Jupyter Notebook files. Their content may differ in the future, so be sure to pull the latest version :)

> Ugly Data > Pandas, Numpy, Dask, ... > Beautiful Data > Machine Learning > ??? > ??? > ??? > Profit