pylearn

Background

A collection of python statistical functions which support rlearn.

Rlearn is an R package for a specific machine learning process. It provides variable selection and linear discriminant analysis (LDA) model fitting and assessment which is linked to the variable selection results. The main output is a list of candidate LDA models.

pylearn builds on the candidate LDA models. It provides some functions for:

• calculating quality assessment scores (user accuracy, producer accuracy, overall accuracy, and the kappa coefficient/Cohen's K)
• munging model quality assessment data with the model candidate list from rlearn
• identifying and removing multicollinear variables given a dataset and a list of candidate variables
• Most notably, pylearn provides a means of predicting relative risk under climate change scenarios by scaling the B_0 coefficient of a logistic regression model by the ratio of intensity-duration-frequency curves.

There are some idiosyncracies to the processes and expected inputs as this was developed for a very specific need - don't hesitate to open an issue if you'd like to start using pylearn but aren't sure where to begin!

Usage

pylearn builds on rlearn, as such you will probably want to get started by using rlearn to build some candidate LDA models. See here to get started with rlearn.

Once you are comfortable with rlearn, there's two options for installing pylearn: Pip or docker.

Installing with Pip

If you are familiar with pip, install pylearn as shown below:

$ pip install git+https://github.com/tesera/pylearn.git@master

Docker is the recommended means of installing and using pylearn if you are not familiar with tools like pip and virtualenv.

Installing with Docker

Docker is the recommended method of using pylearn if you are not fluent with python packaging and isolation tools. Guides for installing and configuring docker can be found for Linux, OSX, and Windows at the docker site. Once you have docker running, install pylearn as shown below:

$ git clone git@github.com/tesera/pylearn
$ cd pylearn
$ docker build -t pylearn .

Using pylearn

Now that you have installed pylearn with pip or docker, start a python session

$ python
# or
$ docker run it pylearn

pylearn will be available for import and usage

import pylearn

Docker Compose

You will need a dev.env file in the project root. The 'dev' container will map the project folder into the container as /opt/pylearn. The test container will run the container, tests and exit.

$ cat dev.env
AWS_ACCESS_KEY_ID=<your-access-key-id>
AWS_SECRET_ACCESS_KEY=<your-secret-key>
AWS_REGION=<your-aws-region>

$ docker-compose run dev

$ docker-compose run test

Contributing

Setup

All contributors are welcome! To get started developing on pylearn we recommend using docker-compose. See docker site to get started with docker-compose.

One you are setup with docker-compose, clone this repo

$ git clone git@github.com:tesera/pylearn.git

You will need a dev.env file in the root project directory using the template below:

$ cat dev.env
AWS_ACCESS_KEY_ID=<your-access-key-id>
AWS_SECRET_ACCESS_KEY=<your-secret-key>
AWS_REGION=<your-aws-region>

Enter the top level directory

cd rlearn
docker-compose run dev

Now you are in the docker container - install the package

pip install --user .

And you're all set to make changes to pylearn!

Testing

Unit tests are required for new functionality. Changes to existing codebase should not break existing tests, or existing tests should be updated if appropriate.

Run the tests as follows

docker-compose run test

Contribution Guidelines

• If you would like to contribute changes to pylearn, please follow this guide to fork, clone, create a branch, make your changes, push your branch to your fork, and open a pull request. Don't forget to run the tests!
• Please follow the Python Style Guide for your contributions to pylearn.

Getting Help

For assistance with usage or development of pylearn, please file an issue on the issue tracker.