Research map / Learning theory and algorithms / Combinatorial and search optimization
HiQ-Lip: A Hierarchical Quantum-Classical Method for Global Lipschitz Constant Estimation of ReLU Networks
Haoqi He, Yan Xiao, Wenzhi Xu, Ruoying Liu, Xiaokai Lin, Kai Wen
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 |
|---|---|---|---|
| use depth-adaptive coefficient (MP B) instead of 1/2^(d-2) (MP A) (ablation) | HiQ-Lip · MNIST Net5 | estimated Lipschitz constant (x10^3): 26.13 -> 1.51 | Improved |
| use depth-adaptive coefficient (MP B) instead of MP A (ablation) | HiQ-Lip · MNIST Net4 | estimated Lipschitz constant (x10^3): 3.25 -> 1.09 | Improved |
| use depth-adaptive coefficient (MP B) instead of MP A (ablation) | HiQ-Lip · MNIST Net3 | estimated Lipschitz constant (x10^3): 0.48 -> 0.48 | No clear effect |
| replace GeoLip with HiQ-Lip | HiQ-Lip · MNIST two-layer, 256 hidden | computation time (s): 99.24 -> 44.79 | Improved |
| replace GeoLip with HiQ-Lip | HiQ-Lip · MNIST two-layer, 8 hidden | computation time (s): 24.07 -> 24.55 | No clear effect |
| replace GeoLip with HiQ-Lip | HiQ-Lip · MNIST two-layer | estimated upper bound: 449.60 -> 448.89 | No clear effect |
| replace LiPopt-2 with HiQ-Lip | HiQ-Lip · MNIST two-layer, 256 hidden | estimated upper bound: 1,088.12 -> 448.89 | Improved |
| replace GeoLip with HiQ-Lip | HiQ-Lip · MNIST Net3 | computation time (s): 784.69 -> 6.46 | Improved |
| replace GeoLip with HiQ-Lip MP B | HiQ-Lip · MNIST Net4/Net5 | estimated Lipschitz constant (x10^3): 0.92 -> 1.09 (Net4), 1.46 -> 1.51 (Net5) | Hurt |
Closest papers by meaning
- ECLipsE: Efficient Compositional Lipschitz Constant Estimation for Deep Neural Networks (NeurIPS 2024 · similarity 0.86)
- TensorRL-QAS: Reinforcement learning with tensor networks for improved quantum architecture search (NeurIPS 2025 · similarity 0.85)
- One More Time: Revisiting Neural Quantum States from a Reinforcement Learning Perspective (NeurIPS 2026 · similarity 0.85)
- PALQO: Physics-informed model for Accelerating Large-scale Quantum Optimization (NeurIPS 2025 · similarity 0.84)
- Characterizing Trainability of Instantaneous Quantum Polynomial Circuit Born Machine (NeurIPS 2026 · similarity 0.84)
- Exponential Hardness of Optimization from the Locality in Quantum Neural Networks (AAAI 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.
