Research map / Learning theory and algorithms / Combinatorial and search optimization
Partition-Based Formulations for Mixed-Integer Optimization of Trained ReLU Neural Networks
Calvin Tsay, Jan Kronqvist, Alexander Thebelt, Ruth Misener
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 |
|---|---|---|---|
| replace big-M with 2-partition formulation | MILP encoding of ReLU NN · MNIST 2x100 optimal adversary | problems solved: 38 -> 59 | Improved |
| replace big-M with 4-partition formulation | MILP encoding of ReLU NN · CIFAR-10 2x100 optimal adversary | problems solved: 62 -> 85 | Improved |
| replace big-M with 2-partition formulation | MILP encoding of ReLU NN · MNIST CNN1 verification eps 0.050 | average solve time (s): 198.5 -> 27.3 | Improved |
| replace big-M with 2 partitions | MILP encoding of ReLU NN · MNIST 2x50 L1-minimal adversary | problems solved: 52 -> 93 | Improved |
| increase partitions from 2 to 4 (ablation) | MILP encoding of ReLU NN · CIFAR-10 CNN2 verification eps 0.010 | problems solved: 72 -> 35 | Hurt |
| increase partitions from 2 to 4 (ablation) | MILP encoding of ReLU NN · MNIST 2x50 optimal adversary eps1=5 | average solve time (s): 42.7 -> 83.9 | Hurt |
| use equal-size or equal-range partitioning instead of random or uneven partitioning (ablation) | partition formulation · MNIST 2x100 optimal adversary | problems solved | Improved |
| add optimization-based bound tightening (ablation) | all formulations · optimal adversary | solve performance | Improved |
Closest papers by meaning
- Optimal Sets and Solution Paths of ReLU Networks (ICML 2023 · similarity 0.90)
- Convex Formulations for Training Two-Layer ReLU Neural Networks (ICLR 2025 · similarity 0.89)
- Convex Relaxations of ReLU Neural Networks Approximate Global Optima in Polynomial Time (ICML 2024 · similarity 0.88)
- Expressivity of ReLU-Networks under Convex Relaxations (ICLR 2024 · similarity 0.87)
- Neural Characteristic Activation Analysis and Geometric Parameterization for ReLU Networks (NeurIPS 2024 · similarity 0.86)
- Compelling ReLU Networks to Exhibit Exponentially Many Linear Regions at Initialization and During Training (ICML 2025 · similarity 0.86)
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.
