Research map / Architectures and efficiency / RNNs and normalization
Biological learning in key-value memory networks
Danil Tyulmankov, Ching Fang, Annapurna Vadaparty, Guangyu Robert Yang
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 |
|---|---|---|---|
| random instead of sequential local third factors (ablation) | key-value memory plasticity rule · Autoassociative recall | capacity: 1.0N -> 0.16N | Hurt |
| key-value memory with sequential third factors instead of classical Hopfield network | autoassociative memory · Autoassociative recall | capacity: 0.14N -> 1.0N | Improved |
| random third-factor probability scales with network size (p=4/N) rather than fixed p=0.1 (ablation) | random local third factor · Autoassociative recall | capacity | Improved |
| replace local-third-factor decay of hidden-to-output synapses with passive decay (meta-learned) (ablation) | meta-learned key-value memory · Autoassociative recall | accuracy | No clear effect |
| key-value memory vs Hopfield network with decay | continual recall · Continual autoassociative recall | accuracy | Improved |
| correlated memories (0.6) instead of uncorrelated | key-value memory plasticity rule · Autoassociative recall | accuracy | No clear effect |
| key-value memory vs BAM on sequence recall | sequence recall · Sequence recall | accuracy | Improved |
| random or trained third factors instead of sequential in copy-paste task (ablation) | copy-paste task · Copy-paste | task performance | Hurt |
Closest papers by meaning
- Long Sequence Hopfield Memory (NeurIPS 2023 · similarity 0.91)
- Large Associative Memory Problem in Neurobiology and Machine Learning (ICLR 2021 · similarity 0.89)
- Universal Hopfield Networks: A General Framework for Single-Shot Associative Memory Models (ICML 2022 · similarity 0.89)
- Memory by accident: a theory of learning as a byproduct of network stabilization (NeurIPS 2025 · similarity 0.89)
- Relating Hopfield Networks to Episodic Control (NeurIPS 2024 · similarity 0.88)
- Predictive Attractor Models (NeurIPS 2024 · similarity 0.88)
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.
