Robust and Provably Monotonic Networks

Abstract

The Lipschitz constant of the map between the input and output space represented by a neural network is a natural metric for assessing the robustness of the model. We present a new method to constrain the Lipschitz constant of dense deep learning models that can also be generalized to other architectures. The method relies on a simple weight normalization scheme during training that ensures the Lipschitz constant of every layer is below an upper limit specified by the analyst. A simple monotonic residual connection can then be used to make the model monotonic in any subset of its inputs, which is useful in scenarios where domain knowledge dictates such dependence. Examples can be found in algorithmic fairness requirements or, as presented here, in the classification of the decays of subatomic particles produced at the CERN Large Hadron Collider. Our normalization is minimally constraining and allows the underlying architecture to maintain higher expressiveness compared to other techniques which aim to either control the Lipschitz constant of the model or ensure its monotonicity. We show how the algorithm was used to train a powerful, robust, and interpretable discriminator for heavy-flavor-quark decays, which has been adopted for use as the primary data-selection algorithm in the LHCb real-time data-processing system in the current LHC data-taking period known as Run 3. In addition, our algorithm has also achieved state-of-the-art performance on benchmarks in medicine, finance, and other applications.

arXiv comment15 pages, 7 figures, v2 extended journal version, v1 presented at the Machine Learning and the Physical Sciences Workshop at the 35th Conference on Neural Information Processing Systems (NeurIPS) December 13, 2021

References in corpus (15)

• Spectral Normalization for Generative Adversarial Networks
• The LHCb Trigger and its Performance in 2011
• Efficient, reliable and fast high-level triggering using a bonsai boosted decision tree
• Invertible Residual Networks
• Design and performance of the LHCb trigger and full real-time reconstruction in Run 2 of the LHC
• LHCb Topological Trigger Reoptimization
• Allen: A high level trigger on GPUs for LHCb
• Searches for dark photons at accelerators
• Sorting out Lipschitz function approximation
• Deep Lattice Networks and Partial Monotonic Functions
• Certified Monotonic Neural Networks
• Pay attention to your loss: understanding misconceptions about 1-Lipschitz neural networks
• Finding NEEMo: Geometric Fitting using Neural Estimation of the Energy Mover's Distance
• A better method to enforce monotonic constraints in regression and classification trees
• Expressive Monotonic Neural Networks

Cited by in corpus (4)

• Non-resonant Anomaly Detection with Background Extrapolation
• Safe but Incalculable: Energy-weighting is not all you need
• Moment Unfolding
• A mathematical certification for positivity conditions in Neural Networks with applications to partial monotonicity and Trustworthy AI