Research map / Probabilistic and scientific modeling / Variational and generative sampling
Expressive probabilistic sampling in recurrent neural networks
Shirui Chen, Linxing Preston Jiang, Rajesh P. N. Rao, Eric Todd SheaBrown
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 reservoir neurons with firing-rate dynamics (RS-FR) to sampler-only network (ablation) | sampler-only network with ReLU · 1-D Gaussian mixture | score function fit / sample match | Improved |
| Use firing-rate instead of synaptic current dynamics in sampler-only network (ablation) | sampler-only network · 1-D Gaussian mixture | score function fit | No clear effect |
| Use tanh instead of ReLU transfer function (ablation) | sampler-only network · 1-D Gaussian mixture with means +-1, variance 0.25 | score function fit | Improved |
| Change mixture to variances 0.0625 and 0.25 (asymmetric score) with tanh | sampler-only network · 1-D Gaussian mixture | score function fit | Hurt |
| Add reservoir neurons (ablation) | reservoir-sampler network · 1-D asymmetric Gaussian mixture, tanh | score function fit | Improved |
| Train RS-FR on 2-D Laplace mixture | RS-FR · 2-D Laplace mixture | density and kurtosis match | Improved |
| Add reservoir neurons with firing-rate dynamics (ablation) | sampler-only networks · MNIST 300-D PCA latent | training loss | Improved |
| Add skew-symmetric divergence-free field J to the drift | reservoir-sampler network · 2-D Laplace mixture | Wasserstein distance over time | Improved |
| Replace linear PCA projection with pretrained nonlinear autoencoder | reservoir-sampler network · MNIST 32-D and CIFAR-10 128-D latents | sample quality | No clear effect |
Closest papers by meaning
- The Bayesian sampling in a canonical recurrent circuit with a diversity of inhibitory interneurons (NeurIPS 2024 · similarity 0.89)
- Complex priors and flexible inference in recurrent circuits with dendritic nonlinearities (ICLR 2024 · similarity 0.89)
- Partial observation can induce mechanistic mismatches in data-constrained models of neural dynamics (NeurIPS 2024 · similarity 0.88)
- MDNS: Masked Diffusion Neural Sampler via Stochastic Optimal Control (NeurIPS 2025 · similarity 0.87)
- Inferring stochastic low-rank recurrent neural networks from neural data (NeurIPS 2024 · similarity 0.87)
- Flow-field inference from neural data using deep recurrent networks (ICML 2025 · similarity 0.87)
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.
