Research map / Probabilistic and scientific modeling / Dynamical systems
Understanding Neural Coding on Latent Manifolds by Sharing Features and Dividing Ensembles
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 hybrid test-time latent optimization rather than the standard variational estimate (ablation) | fae-b · Synthetic circular-latent simulation, N=30 / T=1000 | NLLH: 8830 -> 8501 | Helped |
| Use hybrid test-time latent optimization rather than the standard variational estimate (ablation) | fae-b · Synthetic circular-latent simulation, N=30 / T=1000 | GE: 0.30 -> 0.12 | Helped |
| Share heat-kernel tuning features instead of learning neuron-specific basis weights (ablation) | faeLVM · Synthetic circular-latent simulation, N=30 / T=1000 | GE: 0.17 -> 0.12 | Helped |
| Use a shared heat kernel instead of a shared Gaussian basis (ablation) | faeLVM · Synthetic circular-latent simulation, N=30 / T=1000 | NLLH: 8657 -> 8501 | Helped |
| Remove approximately 50 least informative neurons from each ensemble (ablation) | faeLVM · Rat MEC grid-cell data, modules 2 and 3 | accuracy: 0.83 -> 0.90 | Helped |
| Reduce the encoder convolution kernel size to 1 (ablation) | faeLVM-s · Synthetic circular-latent simulation | GE | Mixed |
| Include temporal transition priors (ablation) | faeLVM · Studied simulation settings | model performance | No clear effect |
| Use the simplified tangent-space KL approximation instead of the specialized variational inference framework (ablation) | faeLVM · Evaluated test cases | model performance | No clear effect |
| Use the simplified tangent-space KL approximation (ablation) | faeLVM · Training | training speed: 5 × | Helped |
| Use feature sharing (ablation) | faeLVM · Synthetic circular-latent simulation, latent recovery | GE | Helped |
| Select the faeLVM-b optimization run with the highest likelihood across five seeds (ablation) | faeLVM-b · Synthetic ensemble detection on two and three rings | ensemble detection accuracy | Helped |
| Use the absolute value of the neural covariance matrix before PCA clustering (ablation) | cov-pca · Synthetic ensemble detection | ensemble detection accuracy | Helped |
Closest papers by meaning
- Meta-Dynamical State Space Models for Integrative Neural Data Analysis (ICLR 2025 · similarity 0.87)
- Connecting Neural Models Latent Geometries with Relative Geodesic Representations (NeurIPS 2025 · similarity 0.86)
- Modeling Latent Neural Dynamics with Gaussian Process Switching Linear Dynamical Systems (NeurIPS 2024 · similarity 0.86)
- CREIMBO: Cross-Regional Ensemble Interactions in Multi-view Brain Observations (ICLR 2025 · similarity 0.86)
- Probabilistic Decomposed Linear Dynamical Systems for Robust Discovery of Latent Neural Dynamics (NeurIPS 2024 · similarity 0.86)
- Multi-modal Gaussian Process Variational Autoencoders for Neural and Behavioral Data (ICLR 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.