Research map / Architectures and efficiency / RNNs and normalization
Tensor decompositions of higher-order correlations by nonlinear Hebbian 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 |
|---|---|---|---|
| Use nonlinear Hebbian learning with a>1 instead of the linear Oja rule | Feedforward neuron learning from higher-order input correlations · Tensor eigenvector learning | convergence target | Mixed |
| Use the dominant tensor eigenvector rather than a subdominant eigenvector as the target | Nonlinear Hebbian learning on natural image patch correlations · Whitened Berkeley image patches | relative basin size: the basin of attraction for U_1 was ~3× larger than that for U_2 | Helped |
| Increase the weight on the two-point correlation term, A_1, with A_2=1 (ablation) | Multi-term generalized Hebbian rule · Whitened natural image patches | alignment to top input-correlation eigenvectors | Helped |
| Weight the three-point correlation more strongly than the two-point correlation (ablation) | Multi-term generalized Hebbian rule · Whitened natural image patches | convergence to the first three-point-correlation eigenvector | Helped |
| Set the two-point correlation weight sufficiently negative (ablation) | Multi-term generalized Hebbian rule · Whitened natural image patches | alignment with input-correlation eigenvectors | Hurt |
| Add a second- and third-order correlation term with shared eigenvectors (ablation) | Multi-term generalized Hebbian rule · Whitened natural image patches | convergence to shared first eigenvector | Helped |
| Add direct multiplicative dependence on synaptic weight (c≠0) (ablation) | Generalized Hebbian learning · Diagonal input correlations | stability of sparse steady states | Mixed |
| Add multiplicative weight dependence with rank-one input correlations (ablation) | Generalized Hebbian learning · Rank-one input correlations | stability of total synaptic amplitude | Hurt |
Closest papers by meaning
- Model Based Inference of Synaptic Plasticity Rules (NeurIPS 2024 · similarity 0.88)
- Spike-timing-dependent Hebbian learning as noisy gradient descent (NeurIPS 2025 · similarity 0.88)
- Neurons as Detectors of Coherent Sets in Sensory Dynamics (NeurIPS 2025 · similarity 0.86)
- Rethinking Hebbian Principle: Low-Dimensional Structural Projection for Unsupervised Learning (NeurIPS 2025 · similarity 0.86)
- Sliding Down the Stairs: How Correlated Latent Variables Accelerate Learning with Neural Networks (ICML 2024 · similarity 0.85)
- Normative Networks for Source Separation via Local Plasticity and Dendritic Computation (NeurIPS 2026 · 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.