CGMVAE: Coupling GMM Prior and GMM Estimator for Unsupervised Clustering and Disentanglement

Impressive progress has been recently witnessed on deep unsupervised clustering and feature disentanglement. In this paper, we propose a novel method top of one recent architecture with explanation Gaussian mixture model (GMM) membership, accompanied by GMM loss to enhance the clustering. The is optimized explicitly computed parameters under our coupled inspired framework. Specifically, takes advantage implicitly learning in latent space neural networks (GMM prior as first GMM), via other framework estimator second GMM). We further introduce Dirichlet conjugate regularization term prevent from degenerating few Gaussians. Eventually, an application apparel generation based proposed which requires only three selection steps. Extensive experiments publicly available datasets demonstrate effectiveness method, terms disentanglement performance.