Reproduce “Mixed data Deep Gaussian Mixture Model: A clustering model for mixed datasets” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Robin Fuchs, Denys Pommeret, Cinzia Viroli
This research introduces the Mixed Deep Gaussian Mixture Model (MDGMM), a model-based clustering approach that unifies Generalized Linear Latent Variable Models and Deep Gaussian Mixture Models to effectively cluster datasets with heterogeneous variable types. The model generates interpretable low-dimensional data representations and, with its robust initialization strategy, consistently achieves strong clustering performance across diverse mixed datasets.
Key insights
- Integrating Generalized Linear Latent Variable Models (GLLVMs) for discrete data and Deep Gaussian Mixture Models (DGMMs) for continuous data into a unified, multi-layered architecture enables effective model-based clustering of mixed datasets.
- The deep architecture provides continuous low-dimensional representations of mixed data, facilitating visualization and revealing complex, non-linear cluster structures.
- A robust initialization strategy (Nested Spaces Embedding Procedure) combined with an automatic model pruning approach significantly improves the stability and usability of EM-based deep mixture models, making complex model selection data-driven.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.