Research map / Probabilistic and scientific modeling / Dynamical systems
Model-based Bayesian inference of neural activity and connectivity from all-optical interrogation of a neural circuit
Laurence Aitchison, Lloyd Russell, Adam M Packer, Jinyao Yan, Philippe Castonguay, Michael Hausser, Srinivas C. Turaga
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 |
|---|---|---|---|
| Use sparse GLM weights instead of dense GLM weights (ablation) | Full model (Sparse GLM + LR) · Spontaneous and perturbed calcium imaging, test ELBO | test ELBO per cell per timestep | Improved |
| Use dense GLM weights instead of only low-rank latents (ablation) | Low-rank model · Spontaneous and perturbed data, test ELBO | test ELBO | No clear effect |
| Joint inference of spikes, low-rank activity and weights instead of a two-stage pipeline (Separate) (ablation) | Full model · Spontaneous and perturbed data, test ELBO | test ELBO | Improved |
| Add optogenetic perturbation data instead of spontaneous data only (>0.95 posterior) | Connectivity inference · Mouse V1 L2/3 connections | number of confident connections: 1204 -> 1940 | Improved |
| Add perturbation data (>=0.99 posterior) | Connectivity inference · Mouse V1 L2/3 connections | number of confident connections: 535 -> 1107 | Improved |
| Add perturbation data (>=0.999 posterior) | Connectivity inference · Mouse V1 L2/3 connections | number of confident connections: 101 -> 527 | Improved |
| Use fresh random stimulation patterns (unfrozen) instead of repeating the same pattern 50 times (frozen) (ablation) | Simulated recovery of weights · Simulated data | correlation with true weights | Improved |
| Reduce noise by a factor of 40 (low-noise) (ablation) | Simulated recovery of weights · Simulated data | correlation with true weights | Improved |
| Additionally reduce calcium transient timescales by a factor of 10 (fast decay) (ablation) | Simulated recovery of weights · Simulated data | correlation with true weights: -> 0.85 | Improved |
| Keep only one low-rank latent | Full model · V1 perturbed data | qualitative latent importance | No clear effect |
Closest papers by meaning
- Fast amortized inference of neural activity from calcium imaging data with variational autoencoders (NeurIPS 2017 · similarity 0.89)
- Active learning of neural population dynamics using two-photon holographic optogenetics (NeurIPS 2024 · similarity 0.88)
- The Bayesian sampling in a canonical recurrent circuit with a diversity of inhibitory interneurons (NeurIPS 2024 · similarity 0.87)
- SpikeSSL: A Universal Spike Inference Framework with Dynamics-Informed State-Space Layers (NeurIPS 2026 · similarity 0.86)
- Scalable inference of functional neural connectivity at submillisecond timescales (NeurIPS 2025 · similarity 0.86)
- Self-Supervised Discovery of Neural Circuits in Spatially Patterned Neural Responses with Graph Neural Networks (NeurIPS 2025 · 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.
