Research map / Probabilistic and scientific modeling / Dynamical systems
Fitting summary statistics of neural data with a differentiable spiking network simulator
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 |
|---|---|---|---|
| Add PSTH and noise-correlation sample-and-measure losses to MLE | RSNN on V1-dataset · V1-dataset test set | Noise correlation R²: -0.91 -> 0.73 | Helped |
| Add PSTH and noise-correlation sample-and-measure losses to MLE | RSNN on V1-dataset · V1-dataset test set | Negative log-likelihood: 0.370 -> 0.370 | No clear effect |
| Fit PSTH only, without noise-correlation matching | RSNN on V1-dataset · V1-dataset test set | Noise correlation R²: -0.91 -> -3.3 | Hurt |
| Add a noise-correlation loss to PSTH fitting | RSNN on V1-dataset · V1-dataset test set | Noise correlation R²: -3.3 -> 0.80 | Helped |
| Add a hidden-neuron population rather than fitting only visible units (ablation) | RSNN on synthetic data · Synthetic dataset, variable hidden-neuron count | Connectivity matrix R²: -0.95 -> 0.50 | Helped |
| Increase the hidden population from 10 to 400 neurons (ablation) | RSNN on synthetic data · Synthetic dataset, variable hidden-neuron count | Connectivity matrix R²: 0.50 -> 0.62 | Helped |
| Increase the hidden population from 10 to 400 neurons (ablation) | RSNN on synthetic data · Synthetic dataset, variable hidden-neuron count | Noise correlation R²: 0.96 -> 0.95 | Mixed |
| Add visible PSTH and noise-correlation losses to ELBO+SM-h (ablation) | RSNN with hidden neurons · Synthetic dataset, 500-neuron target | Noise correlation R²: 0.66±0.1 -> 0.95±0.2 | Helped |
| Add visible PSTH and noise-correlation losses to ELBO+SM-h (ablation) | RSNN with hidden neurons · Synthetic dataset, 500-neuron target | Connectivity matrix R²: 0.56±0.0069 -> 0.63±0.013 | Helped |
| Use sample-and-measure fitting rather than MLE in the misspecified model that omits hidden neurons | RSNN on synthetic data · Synthetic dataset with hidden target neurons and visible-only student | Noise correlation R²: R²=0.95 | Helped |
| Fit with MLE plus PSTH and NC losses in the fully observed, identifiable setting (ablation) | RSNN on synthetic data · Well-specified fully visible synthetic dataset | Connectivity reconstruction | No clear effect |
| Replace the NC loss with the shuffled-trial NC variant (ablation) | RSNN noise-correlation fitting · V1-dataset | Noise-correlation fit | Hurt |
Closest papers by meaning
- Latent Diffusion for Neural Spiking Data (NeurIPS 2024 · similarity 0.87)
- Spike Distance Function as a Learning Objective for Spike Prediction (ICML 2024 · similarity 0.87)
- Causal pieces: analysing and improving spiking neural networks piece by piece (NeurIPS 2026 · similarity 0.86)
- Inferring stochastic low-rank recurrent neural networks from neural data (NeurIPS 2024 · similarity 0.86)
- P-SPIKESSM: HARNESSING PROBABILISTIC SPIKING STATE SPACE MODELS FOR LONG-RANGE DEPENDENCY TASKS (ICLR 2025 · similarity 0.86)
- Scalable inference of functional neural connectivity at submillisecond timescales (NeurIPS 2025 · 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.