Model explainability, interpretability, and reproducibility

Models are often seen as a black box: data goes in, something happens, and a prediction comes out. This lack of transparency is challenging on a number of levels and is often represented in loosely related terms explainability, interpretability, and reproducibility.

  • Explainability: Indicates the description of the internal mechanics of an Machine Learning (ML) model in human terms
  • Interpretability: Indicates the ability to:
    • Understand the relationship between model inputs, features and outputs
    • Predict the response to changes in inputs
  • Reproducibility: Indicates the ability to reproduce the output of a model in a consistent fashion for the same inputs

To solve these challenges, CML provides an end-to-end model governance and monitoring workflow that gives organizations increased visibility into their machine learning workflows and aims to eliminate the blackbox nature of most machine learning models.

The following image shows the end-to-end producltion ML workflow:
Figure 1. Production ML Workflow
Production ML workflow