Pith. sign in

REVIEW 15 cited by

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2012.13349 v3 pith:YECP4NMZ submitted 2020-12-23 math.OC cs.AIcs.DMcs.LGcs.NE

Solving Mixed Integer Programs Using Neural Networks

classification math.OC cs.AIcs.DMcs.LGcs.NE
keywords neuraldatasetsinstanceslearningscipapproachbetterbranching
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
read the original abstract

Mixed Integer Programming (MIP) solvers rely on an array of sophisticated heuristics developed with decades of research to solve large-scale MIP instances encountered in practice. Machine learning offers to automatically construct better heuristics from data by exploiting shared structure among instances in the data. This paper applies learning to the two key sub-tasks of a MIP solver, generating a high-quality joint variable assignment, and bounding the gap in objective value between that assignment and an optimal one. Our approach constructs two corresponding neural network-based components, Neural Diving and Neural Branching, to use in a base MIP solver such as SCIP. Neural Diving learns a deep neural network to generate multiple partial assignments for its integer variables, and the resulting smaller MIPs for un-assigned variables are solved with SCIP to construct high quality joint assignments. Neural Branching learns a deep neural network to make variable selection decisions in branch-and-bound to bound the objective value gap with a small tree. This is done by imitating a new variant of Full Strong Branching we propose that scales to large instances using GPUs. We evaluate our approach on six diverse real-world datasets, including two Google production datasets and MIPLIB, by training separate neural networks on each. Most instances in all the datasets combined have $10^3-10^6$ variables and constraints after presolve, which is significantly larger than previous learning approaches. Comparing solvers with respect to primal-dual gap averaged over a held-out set of instances, the learning-augmented SCIP is 2x to 10x better on all datasets except one on which it is $10^5$x better, at large time limits. To the best of our knowledge, ours is the first learning approach to demonstrate such large improvements over SCIP on both large-scale real-world application datasets and MIPLIB.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 15 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Solving Max-Cut to Global Optimality via Feasibility-Preserving Graph Neural Networks

    cs.LG 2026-05 unverdicted novelty 7.0

    A Max-Cut-specific graph neural network predicts primal- and dual-feasible SDP solutions in linearithmic time, cutting bounding costs in exact branch-and-bound by up to 10.6 times versus a commercial SDP solver while ...

  2. InvEvolve: Evolving White-Box Inventory Policies via Large Language Models with Performance Guarantees

    cs.LG 2026-05 unverdicted novelty 7.0

    InvEvolve evolves white-box inventory policies from LLMs with statistical safety guarantees and outperforms classical and deep learning methods on synthetic and real retail data.

  3. GraphBU: MILP Instance Generation with Graph-Native Block Units

    cs.LG 2026-07 conditional novelty 6.0

    GraphBU generates MILP instances via graph-native block units that pair local subproblems with explicit coupling interfaces, achieving high structural similarity and feasibility preservation across four MILP families.

  4. GRIMIP: A General Framework for Instance-Specific Configuration of MIP Solvers Using LLMs

    cs.LG 2026-06 unverdicted novelty 6.0

    GRIMIP integrates LLMs as probabilistic surrogates inside Bayesian optimization to perform instance-specific MIP solver configuration and reports over 40% reduction in primal-dual integral on hard benchmark instances.

  5. Solving Integer Linear Programming with Parallel Tempering

    cs.LG 2026-05 unverdicted novelty 6.0

    A parallel tempering sampling method with locally-balanced proposals and penalty tempering solves ILP problems competitively with SCIP and Gurobi while showing robustness to distribution shift.

  6. LLM4Branch: Large Language Model for Discovering Efficient Branching Policies of Integer Programs

    cs.AI 2026-05 unverdicted novelty 6.0

    LLM4Branch discovers branching policies for MILP solvers as LLM-generated executable programs whose parameters are tuned via zeroth-order optimization on solver performance.

  7. InvEvolve: Evolving White-Box Inventory Policies via Large Language Models with Performance Guarantees

    cs.LG 2026-05 unverdicted novelty 6.0

    InvEvolve uses LLMs and RL to generate certified inventory policies that outperform classical and deep learning methods on synthetic and real data while providing multi-period performance guarantees.

  8. InvEvolve: Evolving White-Box Inventory Policies via Large Language Models with Performance Guarantees

    cs.LG 2026-05 unverdicted novelty 6.0

    InvEvolve evolves inventory policies using LLMs with RL and provides statistical safety guarantees, outperforming classical and DL methods on synthetic and real data.

  9. A Hybrid Reinforcement and Self-Supervised Learning Aided Benders Decomposition Algorithm

    eess.SY 2026-04 unverdicted novelty 6.0

    A hybrid RL and self-supervised learning method accelerates generalized Benders decomposition by 57.5% on a MINLP case study while recovering optimal solutions.

  10. Feature Augmentation of GNNs for ILPs: Local Uniqueness Suffices

    cs.LG 2025-09 unverdicted novelty 6.0

    Local d-hop uniqueness in GNN node features matches global UID expressiveness for ILP solving while providing stronger generalization.

  11. Unrealized Expectations: Comparing AI Methods vs Classical Algorithms for Maximum Independent Set

    cs.LG 2025-02 unverdicted novelty 6.0

    Classical solver KaMIS outperforms leading AI methods for Maximum Independent Set on random graphs, with some AI approaches no better than simple greedy heuristics and a new serialization analysis revealing similar reasoning.

  12. RL-SPH: Learning to Achieve Feasible Solutions for Integer Linear Programs

    cs.LG 2024-11 unverdicted novelty 6.0

    RL-SPH is a reinforcement learning start primal heuristic that independently produces feasible solutions for ILPs with non-binary integers at 100% rate and with 28.6× lower primal gap than prior start heuristics.

  13. Network Interdiction Goes Neural

    cs.AI 2024-05 unverdicted novelty 6.0

    Multipartite GNN learns MILP formulations of network interdiction to outperform baselines on bi-level combinatorial tasks.

  14. Green Manufacturing Capacity Planning by Integrating Distributionally Robust Optimization and Generative AI

    math.OC 2026-04 unverdicted novelty 5.0

    A DRO model integrated with generative AI for robust capacity planning in green manufacturing under demand and renewable uncertainty.

  15. Feasibility-Aware Imitation Learning for Benders Decomposition

    math.OC 2026-04 unverdicted novelty 5.0

    Feasibility-aware imitation learning accelerates Benders decomposition by predicting feasible integer assignments in the master problem, improving solution times over prior imitation learning methods while retaining f...