Research map / Label-efficient and robust learning / Domain adaptation and generalization
MultiplexNet: Towards Fully Satisfied Logical Constraints in Neural Networks
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Add DNF-based hierarchical super-class constraints to the output layer | Wide ResNet 28-10 on CIFAR100 · CIFAR100 | super-class accuracy: 84.0 \pm (0.2) -> 85.4 \pm (0.3) | Helped |
| Add DNF-based hierarchical super-class constraints to the output layer | Wide ResNet 28-10 on CIFAR100 · CIFAR100 | constraint satisfaction: 83.8 \pm (0.1) -> 100.0 \pm (0.0) | Helped |
| Add DNF-based hierarchical super-class constraints to the output layer | Wide ResNet 28-10 on CIFAR100 · CIFAR100 | class accuracy: 75.0 \pm (0.1) -> 74.4 \pm (0.2) | Hurt |
| Use a hierarchical model that predicts super-class first, then class conditional on it | Wide ResNet 28-10 on CIFAR100 · CIFAR100 | class accuracy: 75.0 \pm (0.1) -> 71.2 \pm (0.2) | Hurt |
| Use a hierarchical model that predicts super-class first, then class conditional on it | Wide ResNet 28-10 on CIFAR100 · CIFAR100 | super-class accuracy: 84.0 \pm (0.2) -> 84.7 \pm (0.1) | Helped |
| Add DL2 constraint loss for hierarchical super-class knowledge | Wide ResNet 28-10 on CIFAR100 · CIFAR100 | class accuracy: 75.0 \pm (0.1) -> 75.3 \pm (0.1) | Helped |
| Add DL2 constraint loss for hierarchical super-class knowledge | Wide ResNet 28-10 on CIFAR100 · CIFAR100 | constraint satisfaction: 83.8 \pm (0.1) -> 85.8 \pm (0.2) | Helped |
| Use arithmetic structure among unlabeled image groups as logical constraints | MultiplexNet on MNIST · MNIST label-free structured learning | inferred-label accuracy: 97.5 \pm 0.3 | Helped |
| Add domain constraints to VAE training | MultiplexNet VAE on synthetic data · Synthetic density estimation | held-out likelihood (-ELBO) | Helped |
| Compile constraints into the VAE output rather than use an unconstrained VAE | MultiplexNet VAE on synthetic data · Synthetic density estimation | constraint satisfaction | Helped |
Closest papers by meaning
- LogicMP: A Neuro-symbolic Approach for Encoding First-order Logic Constraints (ICLR 2024 · similarity 0.84)
- DeepSaDe: Learning Neural Networks That Guarantee Domain Constraint Satisfaction (AAAI 2024 · similarity 0.84)
- HyperLogic: Enhancing Diversity and Accuracy in Rule Learning with HyperNets (NeurIPS 2024 · similarity 0.84)
- Towards Modeling Uncertainties of Self-Explaining Neural Networks via Conformal Prediction (AAAI 2024 · similarity 0.84)
- A2XP: Towards Private Domain Generalization (CVPR 2024 · similarity 0.84)
- Incorporating probabilistic domain knowledge into deep multiple instance learning (ICML 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.