Research map / Continual learning / Online and replay-based
Learning to Learn with Feedback and Local Plasticity
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 |
|---|---|---|---|
| Enable plasticity in non-readout layers instead of using feature reuse only (ablation) | FLP · Incremental Sine Waves, i.i.d. | mean squared error: 0.050 (7e-3) -> 0.00093 (3e-5) | Helped |
| Enable plasticity in non-readout layers instead of using feature reuse only (ablation) | FLP · Incremental Sine Waves, continual | mean squared error: 0.035 (3e-3) -> 0.0016 (6e-5) | Helped |
| Enable plasticity in non-readout layers instead of using feature reuse only (ablation) | FLP · Omniglot | error (%): 4.6 (0.1) -> 3.0 (0.1) | Helped |
| Enable plasticity in non-readout layers instead of using feature reuse only (ablation) | FLP · Omniglot, continual | error (%): 5.0 (0.1) -> 2.7 (0.1) | Helped |
| Enable plasticity in non-readout layers instead of using feature reuse only (ablation) | FLP · Mini-ImageNet | error (%): 48.2 (0.1) -> 42.5 (0.6) | Helped |
| Use 3 rather than 2 plastic layers (ablation) | FLP · Incremental Sine Waves, i.i.d. | mean squared error: 0.00093 (3e-5) -> 0.00051 (8e-5) | Helped |
| Use 3 rather than 2 plastic layers (ablation) | FLP · Incremental Sine Waves, continual | mean squared error: 0.0016 (6e-5) -> 0.0068 (2e-3) | Hurt |
| Use FLP rather than gradient-based inner-loop updates | online meta-learning · Incremental Sine Waves, i.i.d. | mean squared error: 0.00057 (2e-5) -> 0.00051 (8e-5) | No clear effect |
| Use FLP rather than gradient-based inner-loop updates | online meta-learning · Incremental Sine Waves, continual | mean squared error: 0.069 (0.02) -> 0.0016 (6e-5) | Helped |
| Use FLP rather than gradient-based inner-loop updates | online meta-learning · Omniglot | error (%): 3.2 (0.1) -> 3.0 (0.1) | No clear effect |
| Use FLP rather than gradient-based inner-loop updates | online meta-learning · Omniglot, continual | error (%): 3.4 (0.1) -> 2.7 (0.1) | Helped |
| Use FLP rather than gradient-based inner-loop updates | online meta-learning · Mini-ImageNet | error (%): 42.5 (0.3) -> 42.5 (0.6) | No clear effect |
Closest papers by meaning
- A contrastive rule for meta-learning (NeurIPS 2022 · similarity 0.90)
- Learning to solve the credit assignment problem (ICLR 2020 · similarity 0.89)
- Feedback control guides credit assignment in recurrent neural networks (NeurIPS 2024 · similarity 0.89)
- Forward Learning with Top-Down Feedback: Empirical and Analytical Characterization (ICLR 2024 · similarity 0.88)
- Efficient Target Propagation by Deriving Analytical Solution (AAAI 2024 · similarity 0.87)
- Counter-Current Learning: A Biologically Plausible Dual Network Approach for Deep Learning (NeurIPS 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.