From MNIST to ImageNet: Understanding the Scalability Boundaries of Differentiable Logic Gate Networks
Not on the map. Why?
Abstract (arXiv)
Differentiable Logic Gate Networks (DLGNs) are a very fast and energy-efficient alternative to conventional feed-forward networks. With learnable combinations of logical gates, DLGNs enable fast inference by hardware-friendly execution. Since the concept of DLGNs has only recently gained attention, these networks are still in their developmental infancy, including the design and scalability of their output layer. To date, this architecture has primarily been tested on datasets with up to ten classes. This work examines the behavior of DLGNs on large multi-class datasets. We investigate its general expressiveness, its scalability, and evaluate alternative output strategies. Using both synthetic and real-world datasets, we provide key insights into the importance of temperature tuning and its impact on output layer performance. We evaluate conditions under which the Group-Sum layer performs well and how it can be applied to large-scale classification of up to 2000 classes.
Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Builds on 22 of its 33 references that are in the map. First posted on arXiv 30 Sep 2025.
Builds on
22 of 33 references.
References · 22
- 2014ImageNet Large Scale Visual Recognition Challengecited by 8,498
- 2017Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithmscited by 4,349
- 2018DARTS: Differentiable Architecture Searchcited by 1,798
- 2016Neural Architecture Search with Reinforcement Learningcited by 1,499
- 2018Deep Learning for Classical Japanese Literaturecited by 422
- 2017A Downsampled Variant of ImageNet as an Alternative to the CIFAR datasetscited by 377
- 2019Searching for A Robust Neural Architecture in Four GPU Hourscited by 260
- 2019Differentiable Soft Quantization: Bridging Full-Precision and Low-Bit Neural Networkscited by 238
- 2020The Computational Limits of Deep Learningcited by 193
- 2020Binary Neural Networks: A Surveycited by 190
- 2022Deep Differentiable Logic Gate Networkscited by 67
- 2019Cold Case: The Lost MNIST Digitscited by 62
10 more
- 20158-Bit Approximations for Parallelism in Deep Learningcited by 56
- 2024Convolutional Differentiable Logic Gate Networkscited by 37
- 2021Scaling Laws for Deep Learningcited by 26
- 2024Measuring the Energy Consumption and Efficiency of Deep Neural Networks: An Empirical Analysis and Design Recommendationscited by 12
- 2025Recurrent Deep Differentiable Logic Gate Networkscited by 12
- 2020Deep Learning in the Era of Edge Computing: Challenges and Opportunitiescited by 10
- 2025Differentiable Logic Cellular Automata: From Game of Life to Pattern Generationcited by 9
- 2024A Survey on State-of-the-art Deep Learning Applications and Challengescited by 7
- 2021Reduced Precision Strategies for Deep Learning: A High Energy Physics Generative Adversarial Network Use Casecited by 6
- 2023Recent Advancements in Deep Learning Applications and Methods for Autonomous Navigation: A Comprehensive Reviewcited by 6
Built on by
Citing papers per year: 2026: 1.
1 paper.