Generative Kernel Spectral Clustering

TL;DR

Generative Kernel Spectral Clustering (GenKSC) combines spectral clustering with generative modeling to produce interpretable, well-defined clusters and visualizable latent representations, demonstrated on MNIST datasets.

Contribution

The paper introduces GenKSC, a novel model that integrates kernel spectral clustering with generative modeling for improved interpretability and cluster visualization.

Findings

Successfully clusters MNIST and FashionMNIST datasets

Creates explorable latent spaces with visualizable cluster directions

Balances clustering accuracy with interpretability

Abstract

Modern clustering approaches often trade interpretability for performance, particularly in deep learning-based methods. We present Generative Kernel Spectral Clustering (GenKSC), a novel model combining kernel spectral clustering with generative modeling to produce both well-defined clusters and interpretable representations. By augmenting weighted variance maximization with reconstruction and clustering losses, our model creates an explorable latent space where cluster characteristics can be visualized through traversals along cluster directions. Results on MNIST and FashionMNIST datasets demonstrate the model's ability to learn meaningful cluster representations.