active-learning

Example

Sample code in AL_Notebook.ipynb notebook

Install

• pip install active_learning
• OR
• python setup.py sdist
• python setup.py install

Environment Setup

• Make sure that conda is installed.
• Run the following command in the root directory to build the conda environment "trews": conda env create -f environment.yml
• Run source activate trews before executing the Jupyter notebook.
• "trews" has a package nb_conda which allows you to specify the conda env you want as a Jupyter kernel. You must have "trews" activated for the Jupyter notebook server to manage conda environments

Objective

• Provide justification for custom sepsis definition generated from soliciting rich feedback from physicians.
• Create an AL implementation that incorporates rich feedback from physicians to improve the TREWS tool.
• Long-term goal: Create a library of active learning tools for any CDSS we design for new clinical problems.

Code Organization

• General functions should be written in .py files under folder 'python_scripts'
• Experiments should load these python files into iPython notebooks for visualization/output neatness and reproducibility.
  • Large experimental datasets should be stored locally and tracked using plaintext files or logs.
• Datasets used for testing implementation should be kept under repository 'dev_data'