Research map / Probabilistic and scientific modeling / Variational and generative sampling
Automatically Selecting Inference Algorithms for Discrete Energy Minimisation
Paul Henderson, Vittorio Ferrari
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 |
|---|---|---|---|
| learn algorithm selection from GM features (GF) instead of always using the most-often-best algorithm (NB) | inference algorithm selection · OpenGM2 extended, good-and-fastest | mean % matching variables: 75.3 -> 96.4 | Improved |
| learn selection (GF) instead of superclass-based strong baseline (SB) | inference algorithm selection · OpenGM2 extended, good-and-fastest | mean % matching variables: 87.5 -> 96.4 | Improved |
| learn selection (BF) instead of naive baseline NB | inference algorithm selection · OpenGM2 extended, best-and-fastest | % instances correctly classified: 30 -> 62 | Improved |
| learn selection (BF) instead of strong baseline SB | inference algorithm selection · OpenGM2 extended, best-and-fastest | mean % matching variables: 95.6 -> 97.1 | Improved |
| learn selection (BF) instead of strong baseline SB | inference algorithm selection · OpenGM2 extended, best-and-fastest | % instances correctly classified: 36 -> 62 | Improved |
| select algorithm with GF instead of running all algorithms | inference algorithm selection · OpenGM2 extended, efficiency | mean time (s): 13046.8 -> 221.3 | Improved |
| select algorithm with BF instead of running all algorithms | inference algorithm selection · OpenGM2 extended, efficiency | mean time (s): 13046.8 -> 312.5 | Improved |
| use GF (good-and-fastest target) instead of BF (best-and-fastest target) | inference algorithm selection · OpenGM2 extended, efficiency | mean time (s): 312.5 -> 221.3 | Improved |
| evaluate in leave-one-class-out regime instead of random split | GF · OpenGM2 extended, good-and-fastest | mean % matching variables: 96.4 -> 89.8 | Hurt |
| GF versus baselines in LOCO regime | GF · OpenGM2 extended LOCO, good-and-fastest | mean % matching variables: 73.3 -> 89.8 | Improved |
| BF versus strong baseline in LOCO regime | BF · OpenGM2 extended LOCO, best-and-fastest | mean % matching variables: 86.2 -> 85.5 | No clear effect |
| BF versus naive baseline in LOCO regime | BF · OpenGM2 extended LOCO, best-and-fastest | mean % matching variables: 73.3 -> 85.5 | Improved |
Closest papers by meaning
- Complexity of Discrete Energy Minimization Problems (ECCV 2016 · similarity 0.90)
- Partial Optimality by Pruning for MAP-Inference with General Graphical Models (TPAMI 2016 · similarity 0.89)
- Maximum Persistency via Iterative Relaxed Inference in Graphical Models (TPAMI 2017 · similarity 0.88)
- Graph Cuts Always Find a Global Optimum for Potts Models (With a Catch) (ICML 2021 · similarity 0.87)
- MAP Inference via Block-Coordinate Frank-Wolfe Algorithm (CVPR 2019 · similarity 0.86)
- Generalizable Reasoning through Compositional Energy Minimization (NeurIPS 2025 · similarity 0.84)
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.
