Research map / Learning theory and algorithms / Prediction and decision losses
Robust Group Synchronization via Quadratic Programming
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 LP corruption estimation with QP (DESC) formulation | corruption estimation on synthetic UCM data (sigma=0) · synthetic SO(3), n=100, p=0.5 | mean absolute corruption error | Helped |
| Replace LP corruption estimation with QP (DESC) formulation | rotation estimation on synthetic data · synthetic SO(3), sigma=0 | log mean rotation error | Helped |
| Refine DESC-init with IRLS using DESC-estimated corruption levels blended with residuals (ablation) | rotation averaging on Photo Tourism · Photo Tourism (13 datasets) | mean and median rotation error: mean 3.8354 -> 3.5119; median 1.8516 -> 1.5938 | Helped |
| Use DESC-based GCW initialization instead of CEMP-GCW (ablation) | rotation initialization on synthetic data and Photo Tourism · Photo Tourism and synthetic SO(3) | rotation error: Photo Tourism median 2.0339 (CEMP-GCW) -> 1.8516 (DESC-init) | Helped |
| Replace CEMP corruption estimation with DESC QP | corruption estimation on Photo Tourism · Photo Tourism (13 datasets) | median corruption error: median error more than 50% lower on six datasets | Helped |
| Replace CEMP with DESC for corruption estimation | corruption estimation on synthetic data · synthetic SO(3) | mean corruption error | Mixed |
| Replace CEMP with DESC for corruption estimation | corruption estimation on synthetic data · synthetic SO(3), sigma=0 | median corruption error: several orders of magnitude more accurate | Helped |
| DESC runtime per iteration vs CEMP (more iterations needed) | runtime · synthetic data | per-iteration runtime (s): 0.06 s vs 0.02 s per iteration; slower overall | Hurt |
Closest papers by meaning
- Efficient Detection of Long Consistent Cycles and its Application to Distributed Synchronization (CVPR 2024 · similarity 0.92)
- Message Passing Least Squares Framework and its Application to Rotation Synchronization (ICML 2020 · similarity 0.91)
- Robust Angular Synchronization via Directed Graph Neural Networks (ICLR 2024 · similarity 0.85)
- Do We Need Asynchronous SGD? On the Near-Optimality of Synchronous Methods (NeurIPS 2026 · similarity 0.80)
- Equivariant Frames and the Impossibility of Continuous Canonicalization (ICML 2024 · similarity 0.80)
- A Quadratic Synchronization Rule for Distributed Deep Learning (ICLR 2024 · similarity 0.80)
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.