Research map / Architectures and efficiency / RNNs and normalization
Revisit Fuzzy Neural Network: Demystifying Batch Normalization and ReLU with Generalized Hamming Network
Lixin Fan
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 |
|---|---|---|---|
| Add analytical bias rectification based on generalized Hamming distance | GHN · MNIST classification | test accuracy: 99.0% | Improved |
| Add double-thresholding with optimized threshold ratio (ablation) | GHN · CIFAR10 | accuracy: 84.63% -> 89.26% | Improved |
| Add double-thresholding with optimized threshold ratio (ablation) | GHN · CIFAR100 | accuracy: 51.71% -> 60.05% | Improved |
| Disable double-thresholding (ablation) | GHN · MNIST | test accuracy: 98.97%, 98.86%, 98.96%, 98.69% | No clear effect |
| Use an optimized threshold ratio instead of setting r to 0 (ablation) | GHN · CIFAR10 and CIFAR100 | iterations to reach the same test accuracy: 1 to 2 orders of magnitudes | Improved |
| Use analytical bias rectification rather than the baseline VAE setup | MNIST VAE · MNIST reconstruction | reconstruction error: about 28% reduction | Improved |
| Use analytical bias rectification | Sentence CNN · Movie Review sentence classification | evaluation accuracy: both networks yielded the final evaluation accuracy around 74% | No clear effect |
| Use analytical bias rectification | Sentence CNN · Movie Review sentence classification | learning speed | Improved |
| Approximate the input-dependent bias with batch normalization | neural network · MNIST first convolution layer | correlation of neuron outputs with GHD outputs: Pearson correlation coefficient=0.97 | Improved |
| Use a large learning rate of 0.1 rather than 0.01 | GHN · MNIST | learning speed | Improved |
Closest papers by meaning
- Optimization Theory for ReLU Neural Networks Trained with Normalization Layers (ICML 2020 · similarity 0.86)
- Generalized Batch Normalization: Towards Accelerating Deep Neural Networks (AAAI 2019 · similarity 0.86)
- Simplicity Bias and Optimization Threshold in Two-Layer ReLU Networks (ICML 2025 · similarity 0.84)
- Neural Characteristic Activation Analysis and Geometric Parameterization for ReLU Networks (NeurIPS 2024 · similarity 0.84)
- Benign Overfitting and Grokking in ReLU Networks for XOR Cluster Data (ICLR 2024 · similarity 0.84)
- The Effect of Weight Precision on the Neuron Count in Deep ReLU 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.
