Research map / Graph learning / Graph algorithms and clustering
Moccasin: Efficient Tensor Rematerialization for Neural Networks
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use Moccasin's retention-interval CP formulation instead of Checkmate MILP | Memory-constrained rematerialization optimization · Mid-sized graphs where both methods return feasible solutions | solve time: up to an order-of-magnitude less solve time | Helped |
| Use Moccasin instead of Checkmate MILP | RL G2, M=132,156 · RL G2, M=132,156 | TDI (%): 0.9 -> 0.9 | No clear effect |
| Use Moccasin instead of Checkmate MILP | RL G2, M=132,156 · RL G2, M=132,156 | solve time: 685.1 -> 55.0 | Helped |
| Use LP relaxation followed by rounding instead of the full Checkmate MILP | Checkmate rematerialization optimization · RL G2, M=132,156 | TDI (%): 0.9 -> 93.0 | Hurt |
| Use Moccasin instead of Checkmate MILP | Random layered and real-world graphs with more than 250 nodes · Large graphs | finding a feasible solution | Helped |
| Use LP relaxation followed by rounding | Checkmate rematerialization optimization · Most evaluated graphs | memory-budget feasibility | Hurt |
| Limit each node to at most 2 rematerializations/retention intervals (ablation) | Moccasin · Evaluated computation graphs | solution quality | No clear effect |
Closest papers by meaning
- HiRemate: Hierarchical Approach for Efficient Re-materialization of Neural Networks (ICML 2025 · similarity 0.85)
- DOGE-Train: Discrete Optimization on GPU with End-to-End Training (AAAI 2024 · similarity 0.84)
- Position: Tensor Networks are a Valuable Asset for Green AI (ICML 2024 · similarity 0.84)
- Are Graph Neural Networks Optimal Approximation Algorithms? (NeurIPS 2024 · similarity 0.84)
- AT-SKM-Net: An Accelerated Trainable Sampling Kaczmarz-Motzkin Framework for Linear Hard-Constraint Feasibility on Dynamic Graphs (NeurIPS 2026 · similarity 0.84)
- Geometry-aware training of factorized layers in tensor Tucker format (NeurIPS 2024 · 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.