MLDebugger is a framework for finding root causes of errors in machine learning pipelines. For more detailed information about the framework, please refer to our DEEM paper:

Debugging Machine Learning Pipelines. R. Lourenco, J. Freire, and D. Shasha. In Proceedings of the 3rd International Workshop on Data Management for End-to-End Machine Learning (DEEM),2019

The team includes:

• Raoni Lourenço (New York University)
• Juliana Freire (New York University)
• Dennis Shasha (New York University)

We strongly suggest users to read our paper before using our code.

• 1. How To Build
• 2. How To Run

To install latest development version:

$ pip install -e .

To run our example with a machine learning pipeline written in VisTrail, first you need to start a worker:

$ worker &

Then run the executable passing the pipeline path and property-value search space:

$ mldebugger examples/classification_pipeline.vt examples/params.json

More detailed documentation is coming soon.