AI Paper Atlas

Research map / Learning theory and algorithms / Combinatorial and search optimization

MAP Inference Via $$\ell _2$$-Sphere Linear Program Reformulation

Baoyuan Wu, Li Shen, Tong Zhang, Bernard Ghanem

IJCV 2020 · Learning theory and algorithms › Combinatorial and search optimization · approach: Predict-then-optimize learning · 5 citations (OpenAlex)

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 changedWhereMeasuredEffect
add the l2-sphere constraint to the LP relaxation (LS-LP) instead of plain LP relaxation methodsLP-relaxation MAP inference · Gridssolution type: LP-LP, PSDD and AD3 produce uniform solutions on all models; LS-LP gives valid solutionsImproved
use LS-LP instead of l2-Box ADMMMAP inference · Seg-21mean logPot: -344.51 -> -324.89Improved
use LS-LP instead of l2-Box ADMMMAP inference · Grids M14logPot: 44397.5 -> 48766.1Improved
use LS-LP instead of MPLP-CMAP inference · Grids M14logPot: 45813.6 -> 48766.1Improved
use LS-LP instead of MPLP-CMAP inference · Grids M1logPot: 3015.7 -> 2931.8Hurt
use LS-LP instead of MPLP-CMAP inference · Protein M1logPot: -30181.3 -> -30165.5Improved
use LS-LP instead of ogm-ICMMAP inference · Protein M5logPot: -256489 -> -267814Hurt
use LS-LP instead of LP-LP (active-set LP solver)MAP inference · Seg-21runtime (s): 45.71 -> 342Hurt
use ogm-ICM instead of LP-LPMAP inference · Seg-2mean logPot: -75.5 -> -137.1Hurt
add a perturbation epsilon to the objective and constraint of the ADMM (ablation)LS-LP · Convergence analysisglobal convergence guaranteeImproved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy