Research map / Architectures and efficiency / RNNs and normalization
Activation Sharing with Asymmetric Paths Solves Weight Transport Problem without Bidirectional Connection
Sunghyeon Woo, Jeongwoo Park, Jiwoo Hong, Dongsuk Jeon
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 shortcut connections (ablation) | ASAP ResNet-18 · SVHN | test accuracy (%): 93.17 -> 94.86 | Improved |
| add shortcut connections (ablation) | ASAP ResNet-18 · CIFAR-10 | test accuracy (%): 83.44 -> 92.19 | Improved |
| add shortcut connections (ablation) | ASAP ResNet-18 · CIFAR-100 | test accuracy (%): 51.72 -> 68.86 | Improved |
| add shortcut connections (ablation) | BP ResNet-18 · CIFAR-10 | test accuracy (%): 94.70 -> 94.93 | Improved |
| add shortcut connections (ablation) | FA ResNet-18 · CIFAR-10 | test accuracy (%): 71.38 -> 72.47 | Improved |
| increase block size from 2 to 4 (ablation) | ASAP ResNet-34 · CIFAR-10 | test accuracy (%): 93.97 -> 93.9 | No clear effect |
| increase block size from 2 to 4 (ablation) | ASAP ResNet-34 · CIFAR-100 | test accuracy (%): 72.81 -> 71.93 | Hurt |
| increase block size from 2 to 4 (ablation) | ASAP ResNet-34 · memory access | activations used vs BP: 25% | Improved |
| use ASAP instead of FA | ResNet-34 · CIFAR-100 | test accuracy (%): 33.01 -> 72.81 | Improved |
| use ASAP instead of DFA | ResNet-34 · Tiny ImageNet | test accuracy (%): 21.1 -> 52.25 | Improved |
| use ASAP instead of FA | AlexNet · MNIST | test accuracy (%): 99.14 -> 99.32 | No clear effect |
| use bidirectional KP instead of ASAP | ResNet-34 · CIFAR-100 | test accuracy (%): 72.81 -> 75.0 | Improved |
Closest papers by meaning
- Improving equilibrium propagation without weight symmetry through Jacobian homeostasis (ICLR 2024 · similarity 0.85)
- One-Time Soft Alignment Enables Resilient Learning without Weight Transport (NeurIPS 2026 · similarity 0.85)
- Forward Learning with Top-Down Feedback: Empirical and Analytical Characterization (ICLR 2024 · similarity 0.85)
- Layer Collaboration in the Forward-Forward Algorithm (AAAI 2024 · similarity 0.85)
- Improving Deep Learning Speed and Performance Through Synaptic Neural Balance (AAAI 2025 · similarity 0.85)
- Efficient Target Propagation by Deriving Analytical Solution (AAAI 2024 · 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.
