Research map / Learning theory and algorithms / Clustering
EM Algorithms for Weighted-Data Clustering with Application to Audio-Visual Scene Analysis
Israel D. Gebru, Xavier Alameda-Pineda, Florence Forbes, Radu Horaud
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Model data weights as Gamma random variables (WD-EM) instead of fixed weights (FWD-EM) (ablation) | FWD-EM · WAV | Davies-Bouldin index: 1.019 -> 0.975 | Improved |
| Model data weights as random variables (WD-EM) instead of fixed (ablation) | FWD-EM · BCW | Davies-Bouldin index: 0.687 -> 0.622 | Improved |
| Model data weights as random variables (WD-EM) instead of fixed (ablation) | FWD-EM · MNIST | Davies-Bouldin index: 3.104 -> 2.965 | Improved |
| Model weights as random variables (WD-EM) instead of fixed (ablation) | FWD-EM · SIM-Easy (50% outliers) | Davies-Bouldin index: 0.455 -> 0.380 | Improved |
| Model weights as random variables (WD-EM) instead of fixed (ablation) | FWD-EM · SIM-Unbalanced (50% outliers) | Davies-Bouldin index: 2.400 -> 0.557 | Improved |
| Use WD-EM instead of standard GMM | GMM · WAV | Davies-Bouldin index: 1.448 -> 0.975 | Improved |
| Use WD-EM instead of GMM | GMM · SIM-Easy (50% outliers) | Davies-Bouldin index: 0.459 -> 0.380 | Improved |
| Use WD-EM instead of GMM with uniform outlier component | GMM+U · SIM-Easy (50% outliers) | Davies-Bouldin index: 0.195 -> 0.380 | Hurt |
| Use WD-EM instead of I2GMM | I2GMM · MNIST | Davies-Bouldin index: 2.430 -> 2.965 | Hurt |
| Use WD-EM instead of K-Means | K-Means · Letter Recognition | Davies-Bouldin index: 1.450 -> 1.690 | Hurt |
| Use WD-EM instead of GMM+U | GMM-U · Audio-visual moving speakers (MS) | correct detection rate (%): 61.90% -> 83.87% | Improved |
| Use WD-EM instead of FM-uMST | FM-uMST · Audio-visual cocktail party (CP) | correct detection rate (%): 49.57% -> 65.66% | Improved |
Closest papers by meaning
- Gaussian Mixture Models with Rare Events (JMLR 2024 · similarity 0.86)
- A Variational EM Acceleration for Efficient Clustering at Very Large Scales (TPAMI 2021 · similarity 0.85)
- A Flexible EM-Like Clustering Algorithm for Noisy Data (TPAMI 2023 · similarity 0.85)
- Toward Global Convergence of Gradient EM for Over-Paramterized Gaussian Mixture Models (NeurIPS 2024 · similarity 0.83)
- Big Learning Expectation Maximization (AAAI 2024 · similarity 0.83)
- Achieving Optimal Clustering in Gaussian Mixture Models with Anisotropic Covariance Structures (NeurIPS 2024 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
