Lumen is a framework that makes probabilistic models and it's data easily accessible. With its interactive web-frontend users can explore and compare a growing range of probabilistic models, and validate them with their training data - without writing a single line of code. Lumen also allows for easy integration of probabilistic models into other software, as it provides a webservice and a Python API for PQL queries. PQL is Lumen`s SQL-inspired declarative query language for probabilistic models. One of its key features is the support of mixed distributions, i.e. distribution that model not only quantitative but also categorical data attributes.
Lumen is built modular and easily allows the integration of new model types. Currently it supports:
Lumen is free and open-source! Visit our github page.
To our group Lumen also is a playground for other research in the area of visualization and interaction. Setting its model exploration capabilities aside, it's a full fledged interactive data visualization and exploration tool. We use it to try out and evaluate new visualization and interaction ideas.