Supervised learning example: Census population and LANDSAT

[PeterDSteinberg]

This is an example of both datashader and elm. A few things to run this:

• Copy this new elm/examples/LANDSAT-Population-Model.ipynb to datashader/examples
• Modify the yaml of the datashader examples to include elm and earthio and Python 3.5 for now:

  $ cat environment.yml

  
  name: ds
  channels:

• bokeh/label/dev
• conda-forge
• ioam/label/dev
• ioam
• numba
• scitools/label/dev
• elm
• elm/label/dev

dependencies:

• attrs

• bokeh>=0.12.6dev3

• cartopy

• colorcet

• dask

• datashader

• dill

• distributed

• earthio

• elm

• pyarrow

• fastparquet

• geoviews

• holoviews

• iris

• jupyter

• krb5

• matplotlib

• networkx

• numba

• numpy

• pandas

• paramnb

• pyproj

• pytables

• pytest

• python=3.5

• rasterio

• requests

• scikit-image

• scipy

• shapely

• statsmodels

• tblib

• xarray

• yaml

• snappy

• python-snappy

• jupyter_dashboards

• pip:

  • cachey

• Then run the datashader example set up of creating the environment above and running download_example_data.py

• jupyter notebook from the datashader/examples dir

[gbrener]

Hey @PeterDSteinberg , I added a script to this PR which pulls down the necessary files from datashader's examples directory to download the data files and environment.yml. It also patches the environment.yml to look like the one you mentioned in the opening comment of this PR.

This should allow us to not need to clone down the datashader repo and copy files back and forth to get the example notebooks running.

[gbrener]

Also added a README for the [elm/examples]() directory, describing instructions on environment setup and configuring the Jupyter server for the example notebooks. This should address 2 out of 3 checkboxes in Issue #175 .

[PeterDSteinberg]

TODO (me):

• [ ] Provide docs on how to use @gbrener 's helper scripts for environment with:
  • [ ] A note that yaml and requests must be installed first
  • [ ] A note on conda env update if you already have the env or want to name it differently
  • [ ] Viz customization - labels, units, etc
  • [ ] More explanation in notebook
  • [ ] Revised approach to regridding based on recent work in datashader