Bias Variance Agent

Biva library serves as tool to handle bias-variance tradeoff while solving supervised learning problems. Agent provides API-independent approach for model exploration, evaluation and hyperparameter optimisation.

Read the documentation at dsrg.fh-zwickau.de.
Biva is compatible with: Python 3.8.

Getting Started

To get started with the sources clone this git repository to any biva_folder with

git clone https://dsrg.fh-zwickau.de/research/bias-variance-agent.git

and follow these steps:

• with pip and venv:
  1. open a pip terminal in biva_folder. You can make sure that pip is up-to-date by running:
     python -m pip install --upgrade pip
     virtualenv is used to manage Python packages for different projects. Install it with:
     python -m pip install --user virtualenv
  2. Create a virtual environment.
     python -m venv .env
  3. Activate the virtual environment.
     .\.env\Scripts\activate
  4. Run pip install -e . to install this library in development mode
• with conda:
  1. Open conda terminal in biva_folder.
  2. Create a conda environment in .env folder.
     conda env create --force --prefix ./.env python=3.8.*
  3. Activate the conda environment.
     conda activate ./.env
  4. Run pip install -e . to install this library in development mode

Contribution guidelines

To contribute to the project please use pull a request. Before sending it do the follwing:

• Run tests using pytest to make sure everything is working.
• Use pylint to ensure code quallity.

License

Currently undefined.

Scientific Supervisor

• Mike Espig ([email protected])

Authors

• Marcel Becker ([email protected])
• Kostiantyn Pysanyi ([email protected])

Acknowledgments

The research was funded by:
Federal Ministry of Education and Research of Germany (BMBF) and
Saxony Ministry of Science and Art (SMWK).