New E2M algorithm enhances tensor-based density estimation

Researchers have developed a new expectation-maximization (EM) algorithm called E$^2$M for tensor-based discrete density estimation. This algorithm addresses challenges with traditional $\alpha$-divergence methods by first minimizing a surrogate objective using Kullback-Leibler (KL) divergence, which allows for standard EM updates. It then employs a tensor many-body approximation in the M-step to achieve simultaneous closed-form parameter updates. The E$^2$M algorithm supports flexible modeling of various low-rank structures, including CP, Tucker, and Tensor Train formats, and has demonstrated comparable convergence to gradient-based methods, robustness to outliers, and superior density estimation performance on synthetic and real datasets. AI

IMPACT Introduces a novel algorithmic approach for density estimation that could improve performance in various machine learning applications.

RANK_REASON The cluster contains an academic paper detailing a new algorithm and its evaluation. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv stat.ML →

• CP format
• E2M: Double Bounded α-Divergence Optimization for Tensor-based Discrete Density Estimation
• Expectation maximization (EM) algorithms using polar symmetries for computed tomography (CT) image reconstruction.
• Kazu Ghalamkari
• Kullback--Leibler (KL) divergence
• Tucker format

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →