Summary of Deep Clustering Using Dirichlet Process Gaussian Mixture and Alpha Jensen-shannon Divergence Clustering Loss, by Kart-leong Lim
Deep Clustering using Dirichlet Process Gaussian Mixture and Alpha Jensen-Shannon Divergence Clustering Loss
by Kart-Leong Lim
First submitted to arxiv on: 12 Dec 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Computer Vision and Pattern Recognition (cs.CV)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This paper addresses the limitations of current autoencoder-based deep clustering methods, which rely on Kullback-Leibler divergence for clustering loss functions and assume prior knowledge of the number of clusters. The authors propose a new approach that uses Jensen-Shannon divergence to overcome the asymmetry issue and introduces an infinite cluster representation using Dirichlet process Gaussian mixture model for joint clustering and model selection in the latent space, dubbed deep model selection. This method allows the number of clusters to vary dynamically during training, eliminating the need for prior knowledge. The authors evaluate their proposed approach on large-class-number datasets such as MIT67 and CIFAR100, achieving convincing results that outperform traditional variational Bayes models and deep clustering methods. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about finding a better way to group similar things together in pictures using computers. Right now, there are some problems with the ways we do this. We want to make sure the groups aren’t too big or too small, but currently we need to know how many groups there should be beforehand. The authors of this paper have come up with a new way to group things that doesn’t require knowing the number of groups ahead of time. They use a special kind of math called Jensen-Shannon divergence and also introduce a new method called deep model selection, which allows the computer to figure out how many groups there should be as it goes along. The authors tested their new method on big datasets and found that it works better than previous methods. |
Keywords
» Artificial intelligence » Autoencoder » Clustering » Latent space » Mixture model
