Fast SDP certification of neural networks: towards large multi-class datasets
Margot Boyer, Clément Rambour, Zacharie Alès, Amélie Lambert
CEDRIC - OC · ISIR · UMA · CEDRIC - OC
math.CO, stat.ML
Submitted: 2026-07-03
Journal ref: ECML PKDD 2026, Sep 2026, Naples, Italy
Code: https://github.com/MargotBoyer/FastSDPCertification
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- Towards Evaluating the Robustness of Neural Networks
- Formal Verification of Piece-Wise Linear Feed-Forward Neural Networks
- Complete Verification via Multi-Neuron Relaxation Guided Branch-and-Bound
- Explaining and Harnessing Adversarial Examples
- Accelerating Low-Rank Factorization-Based Semidefinite Programming Algorithms on GPU
- Neural Network Branch-and-Bound for Neural Network Verification
- Reluplex: An Efficient SMT Solver for Verifying Deep Neural Networks
- SoK: Certified Robustness for Deep Neural Networks
- Solution-aware vs global ReLU selection: partial MILP strikes back for DNN verification
- Lossless Compression of Deep Neural Networks
Related papers
- Exponential random graph models with soft clique constraints
- Two Linear Programs are Better than One
- Exact Enumeration of Phylogenetic Networks: The Tree-Child, Reticulation-Visible and Orchard Hierarchy
- A complete classification of perfect state transfer in Grover walks on abelian Cayley graphs
- Arithmetic Circuits and Neural Networks for Regular Matroids
- On composition and decomposition operations for vector spaces, graphs and matroids