Mixture Models With a Prior on the Number of Components

A methodological paper by Jeffrey W. Miller of Harvard University introduces mixture of finite mixtures (MFM) models for Bayesian inference with an unknown number of components. The work includes real and simulated data, such as high-dimensional gene expression data used to discriminate cancer subtypes. Supplementary materials for the article are available online.

Use Cases

• Clustering high-dimensional data like gene expression profiles based on the described mixture model framework.
• Comparing inference methods for mixture models with an unknown number of components based on the described MFM and DPM properties.
• Simulating data to test Bayesian nonparametric methods based on the model descriptions.
• Developing or adapting Markov chain Monte Carlo samplers for mixture models based on the discussed methodology.

Strengths

• Includes real-world high-dimensional gene expression data for cancer subtype discrimination as mentioned in the description.
• Method is presented with supplementary materials available online, suggesting additional resources.
• Author affiliation with Harvard University provides institutional credibility.

Limitations

• Row count, column definitions, and sample data are unavailable, limiting suitability assessment.
• Last update date is unknown; freshness unverified.
• Column-level documentation is absent; field semantics must be inferred after download.

Provenance

Source

paperswithcode

Collection Method

Likely contains data generated for or used in the associated academic research paper.