Research map / Probabilistic and scientific modeling / Dynamical systems
Trial matching: capturing variability with data-constrained spiking neural networks
Christos Sourmpis, Carl C. H. Petersen, Wulfram Gerstner, Guillaume Bellec
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 |
|---|---|---|---|
| add trial matching loss (vs trial-average loss only) (ablation) | data-constrained RSNN · Mouse tactile detection test trials | trial-matched Pearson correlation: 0.28 ± 0.003 -> 0.48 ± 0.01 | Improved |
| add trial matching loss (vs trial-average loss only) (ablation) | data-constrained RSNN · Mouse tactile detection test trials | trial-averaged PSTH Pearson correlation: 0.30 ± 0.01 -> 0.30 ± 0.01 | No clear effect |
| add trial matching loss (vs neuron loss only) (ablation) | data-constrained RSNN · Artificial two-area dataset | recovery of bimodal A2 response | Improved |
| soft (Sinkhorn) instead of hard (Hungarian) trial matching (ablation) | trial matching loss · Artificial two-area dataset | solution quality and iterations to converge | No clear effect |
| add trial matching loss (ablation) | data-constrained RSNN · Mouse trial type distribution | trial type frequency match to data | Improved |
| add 5-dimensional latent excitability offset to RSNN (ablation) | RSNN with trial matching · Mouse multi-session dataset | final fitting performance | No clear effect |
| add latent excitability offset (ablation) | RSNN with trial matching · Mouse multi-session dataset | optimization speed | Improved |
| trial-matched likelihood loss alone instead of trial matching (ablation) | RSNN · Artificial two-area dataset | recovery of bimodal hit vs miss distribution | Hurt |
| spike-GAN discriminator on raw spikes instead of trial matching (ablation) | RSNN · Artificial two-area dataset | training success | Hurt |
| increase delay between two areas from 20 ms to 200 ms | RSNN vs LFADS · Extended artificial dataset | fit of generated data | Mixed |
Closest papers by meaning
- A Scalable, Causal, and Energy Efficient Framework for Neural Decoding with Spiking Neural Networks (NeurIPS 2025 · similarity 0.87)
- Controlled maximal variability along with reliable performance in recurrent neural networks (NeurIPS 2024 · similarity 0.86)
- Latent Diffusion for Neural Spiking Data (NeurIPS 2024 · similarity 0.86)
- Leveraging unlabelled data for generalizable neural population decoding (NeurIPS 2026 · similarity 0.86)
- Retrospective for the Dynamic Sensorium Competition for predicting large-scale mouse primary visual cortex activity from videos (NeurIPS 2024 · similarity 0.86)
- Towards a "Universal Translator" for Neural Dynamics at Single-Cell, Single-Spike Resolution (NeurIPS 2024 · 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.
