Research map / Learning theory and algorithms / Combinatorial and search optimization
Improving Feasibility via Fast Autoencoder-Based Projections
Maria Chzhen, Priya L. Donti
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 |
|---|---|---|---|
| add phase 2 adversarial latent structuring (ablation) | FAB · Two Moons, quadratic | feasibility rate (%): 82.5 -> 96.8 | Improved |
| add hinge loss (ablation) | FAB · Two Moons, quadratic | feasibility rate (%): 28.4 -> 96.8 | Improved |
| add discriminator (ablation) | FAB · Two Moons, distance minimization | feasibility rate (%): 13.6 -> 88.3 | Improved |
| add discriminator (ablation) | FAB · Two Moons, quadratic | feasibility rate (%): 75.1 -> 96.8 | Improved |
| use 2 decoder networks instead of 1 (ablation) | FAB · Two Moons, quadratic | feasibility rate (%): 96.8 -> 100.0 | Improved |
| use 2 decoder networks instead of 1 (ablation) | FAB · Two Moons, quadratic | optimality gap: 0.85 -> 1.31 | Hurt |
| use 3 decoder networks instead of 2 (ablation) | FAB · Two Moons, quadratic | feasibility rate (%): 100.0 -> 100.0 | No clear effect |
| add discriminator (ablation) | FAB · Two Moons, quadratic | optimality gap: 0.86 -> 0.85 | No clear effect |
| train autoencoder on only 10% of the constraint set (ablation) | FAB · 2D constraint sets | feasibility rate (%) | No clear effect |
| attach FAB projection to PPO | PPO · SafetyPointGoal2-v0 | episode cost: 167.46 -> 24.32 | Improved |
| attach FAB projection to PPO | PPO · SafetyPointGoal2-v0 | episode reward: 13.26 -> -1.12 | Hurt |
| attach FAB projection to PPO | PPO · SafetyPointPush2-v0 | episode cost: 59.86 -> 7.70 | Improved |
Closest papers by meaning
- Amortized Synthesis of Constrained Configurations Using a Differentiable Surrogate (NeurIPS 2021 · similarity 0.86)
- Iterative Amortized Policy Optimization (NeurIPS 2021 · similarity 0.85)
- A Deep Generative Learning Approach for Two-stage Adaptive Robust Optimization (ICLR 2025 · similarity 0.84)
- Projection-Free Online Convex Optimization with Time-Varying Constraints (ICML 2024 · similarity 0.83)
- Amortized Active Causal Induction with Deep Reinforcement Learning (NeurIPS 2024 · similarity 0.83)
- FSNet: Feasibility-Seeking Neural Network for Constrained Optimization with Guarantees (NeurIPS 2025 · similarity 0.83)
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.
