Project roadmap

Our vision is to establish iMML as a leading and reliable library for multi-modal learning across research and applied settings. This roadmap outlines our priorities to broaden algorithmic coverage, improve performance and scalability, strengthen interoperability, and grow a healthy contributor community.

Guiding principles

  • Practical first: prioritize features that unlock common multi‑modal workflows.

  • Consistent API: keep estimators, transformers, and pipelines aligned with the Scikit-learn and Lightning (for deep learning) interfaces.

  • Reproducible and well‑tested: maintain strong test coverage, deterministic options, and reproducible examples.

  • Interoperable by design: accept standard array‑likes (Numpy, Pandas), and offer bridges to popular ecosystems.

Thematic tracks

Expanding the multi-modal learning field

  • Although iMML originated with a focus on incomplete multi-modal data, most methods can also be applied to fully observed datasets. To broaden adoption and diversity, we will encourage contributions of new multi-modal algorithms, even if they do not explicitly handle missing data.

Algorithm diversity

  • Extend the set of supervised learners that operate on incomplete data natively.

  • Add modules for new tasks: regression, time‑series forecasting/classification, survival analysis, etc.

  • Support additional modality types, such as audio, video and graphs, with new algorithms, utility helpers and examples.

Performance and scalability

  • Re-implement wrapped algorithms in native Python where feasible, simplifying dependencies and improving maintainability.

  • Optimize common operations for speed and memory efficiency.

Interoperability and compatibility

  • Provide support for Polars for users who prefer that backend.

  • Maintain backward compatibility and versioned APIs to minimize user disruption.

Documentation, tutorials, and examples

  • Publish new end-to-end tutorials that cover diverse data workflows and highlight the capabilities of iMML.

  • Develop troubleshooting guides for common pitfalls in multi-modal learning and how to diagnose them.

Community and governance

  • Encourage contributions via clear guidelines and well-scoped tasks.

  • Use GitHub Discussions and Issues for proposals; adopt an issue template for substantial changes.

  • Recognize and celebrate contributors in release notes, documentation, and community channels.

How to get involved

Notes

This roadmap is aspirational and community‑driven. Priorities may shift based on user feedback and contributor interest.