polyview overview

polyview is a sklearn-compatible multi-view learning toolkit for data where each sample is represented by several complementary views, such as audio and video, multiple sensor streams, or heterogeneous feature extractors. Examples and workflow walkthroughs are available in Examples.

The library provides:

• native multi-view clustering algorithms such as multi-view K-means, co-training spectral clustering, co-regularized spectral clustering, and multi-view NMF,

• embedding methods including GCCA and MCCA,

• fusion utilities for early fusion, late fusion, and kernel fusion,

• data augmentation tools that turn a single matrix into multiple views via random projections, random subspaces, or multiple kernels,

• a pipeline layer that moves between single-view and multi-view stages in a consistent sklearn-style API.

The documentation is organized by algorithm family so you can move quickly from the overview to the relevant API reference or example workflow.

This library is inspired by mvlearn, which has not been updated since 2020.

• Multi-View Clustering
• Datasets
• Embedd
• Base API
• Augmentation
• Fusion methods
• PolyPipeline
• Examples
• Miscellaneous