Summary of Deep Embedding Clustering Driven by Sample Stability, By Zhanwen Cheng et al.
Deep Embedding Clustering Driven by Sample Stability
by Zhanwen Cheng, Feijiang Li, Jieting Wang, Yuhua Qian
First submitted to arxiv on: 29 Jan 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: None
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 Deep clustering methods have revolutionized the field by jointly optimizing deep representation learning and clustering, leading to improved task performance. Our novel approach, DECS (Deep Embedding Clustering algorithm driven by Sample Stability), eliminates the need for artificially constructed pseudo targets, which require prior knowledge and can be challenging to determine. Instead, we leverage an autoencoder to construct the initial feature space, then learn a cluster-oriented embedding feature constrained by sample stability. This innovative concept aims to explore deterministic relationships between samples and cluster centroids, effectively pulling samples towards their respective clusters while keeping them away from others with high determinacy. We theoretically verify the model’s convergence using Lipschitz continuity, ensuring its validity. Empirically, our method outperforms state-of-the-art clustering approaches on five datasets. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Imagine you’re trying to group similar objects together without knowing what they really look like inside. That’s kind of like what clustering does! The problem is that most methods need some extra information to work well, which can be hard to get. Our new method, DECS, solves this by using a special type of neural network called an autoencoder to create the initial grouping. Then, it adjusts the grouping based on how well the objects fit into their groups. This helps keep similar things together and keeps others apart. We tested our method on five different sets of data and found that it worked better than other popular methods. |
Keywords
* Artificial intelligence * Autoencoder * Clustering * Embedding * Neural network * Representation learning
