The following tutorials demonstrate the effectiveness of clustering algorithms designed specifically for multiview datasets.
- Multi-view KMeans
- Assessing the Conditional Independence Views Requirement of Multi-view KMeans
- Multi-view vs. Single-view KMeans
- Multi-view Spectral Clustering
- Assessing the Conditional Independence Views Requirement of Multi-view Spectral Clustering
- Multi-view vs Single-view Spectral Clustering
- Multi-view Spherical KMeans
- Multi-view vs Single-view Spherical KMeans
- Using the Multi-view Clustering Algorithm to Cluster Data with Multiple Views
- Multi-view Vs Single-view Visualization and Clustering
The following tutorials demonstrate how effectiveness of cotraining in certain multiview scenarios to boost accuracy over single view methods.
Inference on and visualization of multiview data often requires low-dimensional representations of the data, known as embeddings. Below are tutorials for computing such embeddings on multiview data.
- Generalized Canonical Correlation Analysis (GCCA)
- GCCA vs PCA
- Kernel CCA (KCCA)
- Kernel CCA: ICD Method
- Deep CCA (DCCA)
- CCA Variants Comparison
- Multiview Multidimensional Scaling (MVMDS)
- MVMDS vs PCA
- Omnibus Embedding for Multiview Data
- SplitAE Embeddings on multiview MNIST data
- Predicting views using SplitAE
The following tutorials show how to use multi-view decomposition algorithms.
Methods build on top of Matplotlib and Seaborn have been implemented for convenient plotting of multiview data. See examples of such plots on simulated data.
In order to conviently run tools in this package on multview data, data can be simulated or be accessed from the publicly available UCI multiple features dataset using a dataloader in this package.