Research map / Architectures and efficiency / RNNs and normalization
Backpropagation-Free Deep Learning with Recursive Local Representation Alignment
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 |
|---|---|---|---|
| Train with rec-LRA instead of backprop | ResNet · ImageNet | Top-1 error: 28.15 -> 30.78 | Hurt |
| Train with rec-LRA instead of backprop | ResNet · ImageNet | Top-5 error: 9.81 -> 12.04 | Hurt |
| Train with rec-LRA instead of backprop | ResNet · CIFAR-10 | test error: 5.94 -> 6.42 | Hurt |
| Train with rec-LRA instead of backprop | CNN · CIFAR-10 | test error: 30.17 -> 33.50 | Hurt |
| Train with rec-LRA instead of backprop | 5-layer MLP · FMNIST | test error: 12.98 -> 11.87 | Helped |
| Train with rec-LRA instead of backprop | 5-layer MLP · MNIST | test error: 1.48 -> 1.82 | Hurt |
| Use ReLU instead of tanh (ablation) | rec-LRA 5-layer MLP · MNIST | test error: 1.82 -> 2.26 | Hurt |
| Use ReLU instead of tanh (ablation) | rec-LRA 5-layer MLP · FMNIST | test error: 11.87 -> 14.13 | Hurt |
| Use ELU instead of tanh (ablation) | rec-LRA 5-layer MLP · MNIST | test error: 1.82 -> 1.93 | Hurt |
| Use ELU instead of tanh (ablation) | rec-LRA 5-layer MLP · FMNIST | test error: 11.87 -> 13.17 | Hurt |
| Use non-differentiable sign activation instead of tanh (ablation) | rec-LRA 5-layer MLP · MNIST | test error: 1.82 -> 2.33 | Hurt |
| Use non-differentiable sign activation instead of tanh (ablation) | rec-LRA 5-layer MLP · FMNIST | test error: 11.87 -> 14.88 | Hurt |
Closest papers by meaning
- Accelerated training through iterative gradient propagation along the residual path (ICLR 2025 · similarity 0.89)
- Overcoming Rank Collapse in Feedback Alignment (NeurIPS 2026 · similarity 0.88)
- Reducing Fine-Tuning Memory Overhead by Approximate and Memory-Sharing Backpropagation (ICML 2024 · similarity 0.88)
- One-Time Soft Alignment Enables Resilient Learning without Weight Transport (NeurIPS 2026 · similarity 0.87)
- The Price of Locality: Why Forward-Forward Underperforms Backpropagation? (NeurIPS 2026 · similarity 0.87)
- Towards Interpretable Deep Local Learning with Successive Gradient Reconciliation (ICML 2024 · similarity 0.86)
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.