Research map / Architectures and efficiency / RNNs and normalization
Sparsely-Connected Neural Networks: Towards Efficient VLSI Implementation of Deep 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 |
|---|---|---|---|
| Use 50% fewer connections in the SPFP network | MNIST 784-1024-1024-1024-10 · MNIST | misclassification rate (%): 1.33 -> 1.17 | Helped |
| Use 90% fewer connections in the SPFP network | MNIST 784-1024-1024-1024-10 · MNIST | misclassification rate (%): 1.33 -> 1.33 | No clear effect |
| Add 50% connectivity sparsity | BinaryConnect · MNIST | misclassification rate (%): 1.23 -> 0.99 | Helped |
| Add 90% connectivity sparsity | BinaryConnect · MNIST | misclassification rate (%): 1.23 -> 1.33 | Hurt |
| Add 50% connectivity sparsity | TernaryConnect · MNIST | misclassification rate (%): 1.15 -> 0.95 | Helped |
| Add 90% connectivity sparsity | TernaryConnect · MNIST | misclassification rate (%): 1.15 -> 1.41 | Hurt |
| Add 90% connectivity sparsity | SPFP convolutional network · CIFAR10 | misclassification rate (%) without data augmentation: 12.45 -> 12.05 | Helped |
| Add 90% connectivity sparsity | BinaryConnect convolutional network · CIFAR10 | misclassification rate (%) without data augmentation: 9.91 -> 8.05 | Helped |
| Add 90% connectivity sparsity | TernaryConnect convolutional network · CIFAR10 | misclassification rate (%) without data augmentation: 9.32 -> 7.88 | Helped |
| Add 90% connectivity sparsity | BinaryConnect convolutional network · SVHN | misclassification rate (%): 2.134615 -> 2.003846 | Helped |
| Add 90% connectivity sparsity | TernaryConnect convolutional network · SVHN | misclassification rate (%): 2.9 -> 1.957692 | Helped |
| Use LFSR-generated sparse connections at p = 0.9375 | ASIC single neuron · TSMC 65 nm CMOS, 400 MHz | energy (pJ): 712 -> 110 | Helped |
Closest papers by meaning
- DepthShrinker: A New Compression Paradigm Towards Boosting Real-Hardware Efficiency of Compact Neural Networks (ICML 2022 · similarity 0.88)
- Fast Sparse ConvNets (CVPR 2020 · similarity 0.88)
- Random Sparse Lifts: Construction, Analysis and Convergence of finite sparse networks (ICLR 2024 · similarity 0.86)
- Scalable Neural Network Kernels (ICLR 2024 · similarity 0.85)
- On the Hardness of Training Deep Neural Networks Discretely (AAAI 2025 · similarity 0.85)
- Accelerating Linear Recurrent Neural Networks for the Edge with Unstructured Sparsity (ICML 2025 · similarity 0.85)
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.