Research map / Optimization / Nonconvex and smooth optimization
Consensus-Based Optimization on the Sphere: Convergence to Global Minimizers and Machine Learning
Massimo Fornasier, Lorenzo Pareschi, Hui Huang, Philippe Sünnen
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 |
|---|---|---|---|
| Use the fast method with a decreasing number of particles (mu=0.3 or 0.2, N_min=10) | KV-CBO · Ackley d=20 | performance for given computational cost | Improved |
| Increase alpha from 5 to 500 (ablation) | KV-CBO · Ackley d=3 | accuracy of minimum | Improved |
| Use KV-CBO instead of CBO on Euclidean space with sigma=2 | consensus-based optimization · Ackley d=3 | consensus error: success rate 100% for both | No clear effect |
| Increase number of particles N and frame vectors M (ablation) | KV-CBO · Ackley d=20, v* shifted | success rate: 98% -> 99% -> 100% | Improved |
| Set sigma=0 (deterministic Kuramoto-Vicsek) instead of adding noise (ablation) | KV-CBO · Robust subspace d=10, N=2500 | success rate: nearly 60% | Hurt |
| Use more particles (ablation) | KV-CBO · Robust subspace detection | window of usable sigma | Improved |
| Use KV-CBO instead of Wirtinger Flow in phase retrieval with adversarial noise | phase retrieval · d=32, M=4d | success rate vs SNR | No clear effect |
| Use KV-CBO instead of Wirtinger Flow / PhaseMax in phase retrieval phase transition | phase retrieval · d=32 | perfect recovery vs M | Hurt |
| Random subspaces instead of nearly parallel subspaces needs larger N | KV-CBO · Robust subspace p=2 | particles needed | Hurt |
Closest papers by meaning
- Multi-Consensus Decentralized Accelerated Gradient Descent (JMLR 2023 · similarity 0.87)
- On the Convergence of Decentralized Stochastic Minimax Optimization Algorithm with Compressed Communication (ICML 2026 · similarity 0.87)
- Replica Exchange for Non-Convex Optimization (JMLR 2021 · similarity 0.86)
- Distributed Second Order Methods with Fast Rates and Compressed Communication (ICML 2021 · similarity 0.86)
- Variational inference via Gaussian interacting particles in the Bures-Wasserstein geometry (ICML 2026 · similarity 0.86)
- Decentralized Stochastic Optimization and Gossip Algorithms with Compressed Communication (ICML 2019 · similarity 0.86)
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.
