Research map / Optimization / Optimizers
Deep Differentiable Logic Gate Networks
Felix Petersen, Christian Borgelt, Hilde Kuehne, Oliver Deussen
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 differentiable logic gate net instead of neural network at color-channel resolution 4 | CIFAR-10 small model · CIFAR-10 | accuracy: 50.79% -> 51.27% | Improved |
| scale logic net size from small to large x4 | Diff Logic Net · CIFAR-10 | accuracy: 51.27% -> 62.14% | Improved |
| use logic gate net instead of neural network on MONK-2 | MONK-2 · MONK-2 | accuracy: 100% -> 90.9% | Hurt |
| use logic gate net instead of neural network on MONK-3 (label noise) | MONK-3 · MONK-3 | accuracy: 93.5% -> 97.7% | Improved |
| use logic gate net instead of neural network on Adult | Adult · Adult | accuracy: 84.9% -> 84.8% | No clear effect |
| use logic gate net instead of neural network on MNIST | MNIST · MNIST | accuracy: 98.40% -> 98.47% | Improved |
| use logic gate net instead of FINN on MNIST | MNIST · MNIST | BNN comparison: 98.40% -> 98.47%; 5.28M -> 384K ops | Improved |
| discretize relaxed network to hard gates after training | Diff Logic Net · MNIST | accuracy: gap < 0.1% | No clear effect |
| reduce the set of logic operators (ablation) | Diff Logic Net · all settings | accuracy | Hurt |
| use Adam instead of SGD (ablation) | Diff Logic Net training · training | training quality | Improved |
| half precision training instead of float32 (ablation) | Diff Logic Net · training | training performance | No clear effect |
Closest papers by meaning
- Convolutional Differentiable Logic Gate Networks (NeurIPS 2024 · similarity 0.94)
- Light Differentiable Logic Gate Networks (ICLR 2026 · similarity 0.93)
- Mind the Gap: Removing the Discretization Gap in Differentiable Logic Gate Networks (NeurIPS 2025 · similarity 0.92)
- LogicMP: A Neuro-symbolic Approach for Encoding First-order Logic Constraints (ICLR 2024 · similarity 0.85)
- BOLD: Boolean Logic Deep Learning (NeurIPS 2024 · similarity 0.83)
- Differentiable Weightless Neural Networks (ICML 2024 · similarity 0.83)
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.
