Pith. sign in

REVIEW 11 cited by

Evaluating Robustness of Neural Networks with Mixed Integer Programming

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 1711.07356 v3 pith:2KGB6JGP submitted 2017-11-20 cs.LG cs.CRcs.CV

Evaluating Robustness of Neural Networks with Mixed Integer Programming

classification cs.LG cs.CRcs.CV
keywords networksadversarialexamplesneuralaccuracyboundedclassifiercomputational
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
read the original abstract

Neural networks have demonstrated considerable success on a wide variety of real-world problems. However, networks trained only to optimize for training accuracy can often be fooled by adversarial examples - slightly perturbed inputs that are misclassified with high confidence. Verification of networks enables us to gauge their vulnerability to such adversarial examples. We formulate verification of piecewise-linear neural networks as a mixed integer program. On a representative task of finding minimum adversarial distortions, our verifier is two to three orders of magnitude quicker than the state-of-the-art. We achieve this computational speedup via tight formulations for non-linearities, as well as a novel presolve algorithm that makes full use of all information available. The computational speedup allows us to verify properties on convolutional networks with an order of magnitude more ReLUs than networks previously verified by any complete verifier. In particular, we determine for the first time the exact adversarial accuracy of an MNIST classifier to perturbations with bounded $l_\infty$ norm $\epsilon=0.1$: for this classifier, we find an adversarial example for 4.38% of samples, and a certificate of robustness (to perturbations with bounded norm) for the remainder. Across all robust training procedures and network architectures considered, we are able to certify more samples than the state-of-the-art and find more adversarial examples than a strong first-order attack.

discussion (0)

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

Forward citations

Cited by 11 Pith papers

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

  1. ShardNet: Training Neural Controllers with Hard, Non-Convex Constraints

    eess.SY 2026-06 unverdicted novelty 8.0

    ShardNet enforces non-convex polyhedral safety constraints in neural controllers by construction via a differentiable projection layer, achieving 100% verified safety and over 3x larger safe sets than prior methods on...

  2. Stress-Testing Neural Network Verifiers with Provably Robust Instances

    cs.LG 2026-05 conditional novelty 7.0

    A reusable framework generates verification instances with provably known robustness labels, revealing numeric tolerance issues and bugs in five verifiers while introducing difficulty profiles to diagnose failure modes.

  3. Towards Generalized Certified Robustness with Multi-Norm Training

    cs.LG 2024-10 unverdicted novelty 7.0

    CURE is the first multi-norm certified training method that improves union robustness across l_p norms and unseen perturbations on MNIST, CIFAR-10 and TinyImagenet.

  4. Scaling Neural Network Verification with Tensor Parallelism and Fully Sharded Data Parallelism

    cs.LG 2026-06 unverdicted novelty 6.0

    Adapts TP and FSDP to bound-propagation verification, with FSDP delivering bitwise-identical bounds and 80-90% baseline memory reduction while TP trades some tightness for ~2x peak-memory savings.

  5. Neural Network Verification using Partial Multi-Neuron Relaxation

    cs.LO 2026-05 unverdicted novelty 6.0

    Introduces partial multi-neuron relaxation using existing branching heuristics to balance bound tightness and scalability in neural network verification, with integration into Marabou showing positive experimental com...

  6. Neural Two-Stage Stochastic Volt-VAR Optimization for Three-Phase Unbalanced Distribution Systems with Network Reconfiguration

    eess.SY 2025-10 unverdicted novelty 6.0

    A neural network approximates the second-stage recourse model in two-stage stochastic Volt-VAR optimization, allowing the full problem to be solved as a mixed-integer linear program with over 50x speedup and sub-0.3% ...

  7. Neural Two-Stage Stochastic Optimization for Solving Unit Commitment Problem

    eess.SY 2025-07 unverdicted novelty 6.0

    A neural network approximates recourse costs in two-stage stochastic unit commitment, is embedded as an MILP, and uses a scenario-embedding network for scalable solving with low optimality gaps.

  8. Polyhedral Enclosures: An Efficient Combinatorial Abstraction for Nonlinear Neural Feedback Systems

    eess.SY 2025-03 unverdicted novelty 6.0

    Presents a polyhedral enclosure algorithm encoded as MILP to produce sound over-approximations of forward reachable sets for nonlinear neural feedback systems.

  9. A Performance Analyzer for a Public Cloud's ML-Augmented VM Allocator

    cs.DC 2025-12 unverdicted novelty 5.0

    SANJESH applies bi-level optimization to production traces and reveals VM allocation scenarios that cause 4x worse performance than the operator's existing evaluator detected.

  10. veriFIRE: an Industrial Case Study in Verifying Consistency Properties for a DNN-Based Wildfire Detection System

    cs.LO 2026-06 unverdicted novelty 4.0

    An industrial case study applies neural network verifiers to check monotonicity and bounded-response consistency properties in a DNN-based wildfire detection system on real background samples.

  11. Bridging Control with Neural Network Verifier alpha-beta-CROWN: A Tutorial

    eess.SY 2026-05 unverdicted novelty 3.0

    Tutorial introducing applications of the existing α,β-CROWN verifier to scalable formal verification of neural network controllers via bound computation and domain partitioning.