{"title":"Augmented ELBO regularization for enhanced clustering in variational autoencoders","authors":"Kwangtek Na , Ju-Hong Lee , Eunchan Kim","doi":"10.1016/j.neucom.2024.128795","DOIUrl":null,"url":null,"abstract":"<div><div>With significant advances in deep neural networks, various new algorithms have emerged that effectively model latent structures within data, surpassing traditional clustering methods. Each data point is expected to belong to a single cluster in a typical clustering algorithm. However, clustering based on variational autoencoders (VAEs) represents the expectation of the overall clusters, denoted as <span><math><mrow><mi>c</mi><mo>=</mo><mn>1</mn><mo>,</mo><mo>…</mo><mo>,</mo><mi>K</mi></mrow></math></span> in the KL divergence term. Consequently, the latent embedding <span><math><mi>z</mi></math></span> can be learned to exist across multiple clusters with relatively balanced probabilities, rather than being strongly associated with a specific cluster. This study introduces an additional regularizer to encourage the latent embedding <span><math><mi>z</mi></math></span> to have a strong affiliation with specific clusters. We introduce optimization methods to maximize the ELBO that includes the newly added regularization term and explore methods to eliminate computationally challenging terms. The positive impact of this regularization on clustering accuracy was verified by examining the variance of the final cluster probabilities. Furthermore, an enhancement in the clustering performance was observed when regularization was introduced.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":null,"pages":null},"PeriodicalIF":5.5000,"publicationDate":"2024-10-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Neurocomputing","FirstCategoryId":"94","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S0925231224015662","RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE","Score":null,"Total":0}
引用次数: 0
Abstract
With significant advances in deep neural networks, various new algorithms have emerged that effectively model latent structures within data, surpassing traditional clustering methods. Each data point is expected to belong to a single cluster in a typical clustering algorithm. However, clustering based on variational autoencoders (VAEs) represents the expectation of the overall clusters, denoted as in the KL divergence term. Consequently, the latent embedding can be learned to exist across multiple clusters with relatively balanced probabilities, rather than being strongly associated with a specific cluster. This study introduces an additional regularizer to encourage the latent embedding to have a strong affiliation with specific clusters. We introduce optimization methods to maximize the ELBO that includes the newly added regularization term and explore methods to eliminate computationally challenging terms. The positive impact of this regularization on clustering accuracy was verified by examining the variance of the final cluster probabilities. Furthermore, an enhancement in the clustering performance was observed when regularization was introduced.
随着深度神经网络的长足发展,各种新算法应运而生,它们能有效地模拟数据中的潜在结构,超越了传统的聚类方法。在典型的聚类算法中,每个数据点都属于一个聚类。然而,基于变异自编码器(VAE)的聚类代表了整体聚类的期望值,在 KL 发散项中表示为 c=1,...,K。因此,可以学习到潜在嵌入 z 以相对均衡的概率存在于多个聚类中,而不是与特定聚类紧密相关。本研究引入了一个额外的正则因子,以鼓励潜在内嵌 z 与特定聚类有较强的关联。我们引入了优化方法来最大化包含新添加的正则项的 ELBO,并探索了消除计算上具有挑战性的项的方法。通过检查最终聚类概率的方差,验证了正则化对聚类准确性的积极影响。此外,引入正则化后,聚类性能也得到了提高。
期刊介绍:
Neurocomputing publishes articles describing recent fundamental contributions in the field of neurocomputing. Neurocomputing theory, practice and applications are the essential topics being covered.