Bayesian ICA with super-Gaussian Source Priors
Fuente:
arXiv
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866908650743791616 |
|---|---|
| author | Datta, Jyotishka Ghosh, Soham Polson, Nicholas G. |
| author_facet | Datta, Jyotishka Ghosh, Soham Polson, Nicholas G. |
| contents | Independent Component Analysis (ICA) plays a central role in modern machine learning as a flexible framework for feature extraction. We introduce a horseshoe-type prior with a latent Polya-Gamma scale mixture representation, yielding scalable algorithms for both point estimation via expectation-maximization (EM) and full posterior inference via Markov chain Monte Carlo (MCMC). This hierarchical formulation unifies several previously disparate estimation strategies within a single Bayesian framework. We also establish the first theoretical guarantees for hierarchical Bayesian ICA, including posterior contraction and local asymptotic normality results for the unmixing matrix. Comprehensive simulation studies demonstrate that our methods perform competitively with widely used ICA tools. We further discuss implementation of conditional posteriors, envelope-based optimization, and possible extensions to flow-based architectures for nonlinear feature extraction and deep learning. Finally, we outline several promising directions for future work. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2406_17058 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Bayesian ICA with super-Gaussian Source Priors Datta, Jyotishka Ghosh, Soham Polson, Nicholas G. Methodology Machine Learning 62F15, 62H25, 68T07 Independent Component Analysis (ICA) plays a central role in modern machine learning as a flexible framework for feature extraction. We introduce a horseshoe-type prior with a latent Polya-Gamma scale mixture representation, yielding scalable algorithms for both point estimation via expectation-maximization (EM) and full posterior inference via Markov chain Monte Carlo (MCMC). This hierarchical formulation unifies several previously disparate estimation strategies within a single Bayesian framework. We also establish the first theoretical guarantees for hierarchical Bayesian ICA, including posterior contraction and local asymptotic normality results for the unmixing matrix. Comprehensive simulation studies demonstrate that our methods perform competitively with widely used ICA tools. We further discuss implementation of conditional posteriors, envelope-based optimization, and possible extensions to flow-based architectures for nonlinear feature extraction and deep learning. Finally, we outline several promising directions for future work. |
| title | Bayesian ICA with super-Gaussian Source Priors |
| topic | Methodology Machine Learning 62F15, 62H25, 68T07 |
| url | https://arxiv.org/abs/2406.17058 |