Research map / Optimization / Nonconvex and smooth optimization
Generalized Majorization-Minimization
Sobhan Naderi Parizi, Kun He, Reza Aghajani, Stan Sclaroff, Pedro Felzenszwalb
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 |
|---|---|---|---|
| replace the touching-bound requirement with selection among valid bounds | G-MM versus MM on k-means clustering · GMM-200 and Cloud | objective: GMM-200 random partition: 11.20 \pm 0.63 -> 1.85 \pm 0.02; Cloud random partition: 44453 \pm 88341 -> 1470 \pm 8 | Improved |
| replace CCP bound selection with random valid-bound selection | LS-SVM on mammals · Mammals, top-left initialization | test error: 42.5 \pm 4.6 -> 31.4 \pm 10.1 | Improved |
| select bounds using a multi-fold-MIL-inspired bias | G-MM on LS-SVM · Mammals, top-left initialization | test error: 42.5 \pm 4.6 -> 18.9 \pm 5.0 | Improved |
| favor coherent part assignments when selecting bounds | G-MM on LS-SVM · MIT-Indoor, λ=1.00 initialization | accuracy: 46.62 -> 56.03 | Improved |
| use random valid bounds instead of the CCP bound | G-MM on LS-SVM · MIT-Indoor, random initialization | accuracy: 41.94 \pm 1.1 -> 47.51 \pm 0.7 | Improved |
| increase the progress coefficient | G-MM clustering · D31 and Cloud | objective quality across runs | Hurt |
| set the progress coefficient to 1, recovering standard k-means/MM bound selection (ablation) | G-MM clustering · clustering datasets | objective | Hurt |
| use random partition initialization rather than Forgy or k-means++ initialization (ablation) | G-MM clustering · D31 and GMM-200 | objective: GMM-200: 1.85 \pm 0.02; D31: 1.21 \pm 0.05 | Improved |
| use coherent rather than less coherent initial part assignments (ablation) | LS-SVM on MIT-Indoor · MIT-Indoor | accuracy: CCP: 40.88 -> 46.62; G-MM random: 43.38 -> 49.88; G-MM biased: 44.83 -> 56.03 | Mixed |
Closest papers by meaning
- Stochastic Regularized Majorization-Minimization with weakly convex and multi-convex surrogates (JMLR 2024 · similarity 0.89)
- Relaxed Majorization-Minimization for Non-Smooth and Non-Convex Optimization (AAAI 2016 · similarity 0.88)
- Sharper Generalization Bounds for Learning with Gradient-dominated Objective Functions (ICLR 2021 · similarity 0.88)
- Towards Generalization beyond Pointwise Learning: A Unified Information-theoretic Perspective (ICML 2024 · similarity 0.85)
- A Universal Class of Sharpness-Aware Minimization Algorithms (ICML 2024 · similarity 0.85)
- Near-Optimal Solutions of Constrained Learning Problems (ICLR 2024 · similarity 0.85)
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.
