AI Paper Atlas

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

USCO-Solver: Solving Undetermined Stochastic Combinatorial Optimization Problems

Guangmo Tong

NeurIPS 2021 · Learning theory and algorithms › Combinatorial and search optimization · approach: Unsupervised neural combinatorial optimization · 1 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
Increase the number of φ_exp configurations from 16 to 3200 (ablation)USCO-Solver on SSP Col · SSP Colperformance ratio: 1.942 { (0.11)} -> 1.129 { (0.01)}Improved
Use configurations from the true distribution rather than φ_exp (ablation)USCO-Solver on SSP Col · SSP Colperformance ratio: 1.129 { (0.01)} -> 1.010 { (0.01)}Improved
Increase the number of φ_exp configurations from 80 to 6400 (ablation)USCO-Solver on SSP NY · SSP NYperformance ratio: 1.613 { (0.01)} -> 1.192 { (0.01)}Improved
Use Gaussian φ_norm configurations rather than φ_exp configurations (ablation)USCO-Solver on SSP · SSP Kroperformance ratio: 11.29 { (0.34)} vs 5.829 { (0.22)}Hurt
Use USCO-Solver rather than the strongest reported baselineSSP · SSP Colperformance ratio: 1.956 { (0.10)} -> 1.129 { (0.01)}Improved
Increase the number of φ_uni configurations from 8 to 640 (ablation)USCO-Solver on SSC Cora · SSC Coraperformance ratio: 1.480 { (0.12)} -> 1.083 { (0.01)}Improved
Use USCO-Solver with φ_uni rather than GNNSSC · SSC Coraperformance ratio: 1.036 { (0.02)} vs 1.083 { (0.01)}Hurt
Increase the number of φ_uni configurations from 16 to 19200 (ablation)USCO-Solver on SBM · SBMperformance ratio: 3.627 { (0.01)} -> 2.696 { (0.01)}Improved
Use a prior informed by edge-weight means (φ_1) rather than φ_uni (ablation)USCO-Solver on SBM · SBMperformance ratio: 3.582 { (0.01)} -> 1.008 { (<0.01)}Improved
Use a narrow φ_0.3 prior rather than the broader φ_10 prior (ablation)USCO-Solver on SBM · SBMperformance ratio: 4.889 { (0.09)} -> 1.008 { (<0.01)}Improved
Increase the training set while using φ_uni configurations (ablation)USCO-Solver on SBM · SBMperformance ratio: 3.107 { (0.04)} -> 2.966 { (0.02)}Improved
Increase SSP NY training size from 160 to 1600 (ablation)USCO-Solver (φ_exp) · SSP NYperformance ratio: 1.613 { (0.01)} -> 1.989 { (0.03)}Hurt

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