Research map / Probabilistic and scientific modeling / Dynamical systems
Taking the neural sampling code very seriously: A data-driven approach for evaluating generative models of the visual system
Suhas Shrinivasan, Konstantin-Klemens Lurz, Kelli Restivo, George Denfield, Andreas S. Tolias, Edgar Y. Walker, Fabian H. Sinz
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 |
|---|---|---|---|
| replace exponential prior with normalizing-flow prior (variational dequantization) (ablation) | NSC generative model · V1 session with 29 neurons, prior log-likelihood | bits/neuron/trial: 0.095 bits per neuron per trial | Improved |
| replace linear likelihood with MLP likelihood (ablation) | NSC generative model · V1 session with 29 neurons, likelihood log-likelihood | bits/pixel/trial: 0.052 bits per pixel per trial (87.19 bits) | Improved |
| use Flow prior plus MLP likelihood versus Exp1-Lin baseline (ablation) | NSC generative model · V1 session with 29 neurons, joint log-likelihood | bits/pixel/neuron/trial: 1.8452e-3 bits per pixel per neuron per trial (89.951 bits) | Improved |
| use flexible generative model instead of Exp1-Lin baseline (ablation) | NSC posterior · V1 session with 29 neurons, posterior log-likelihood | bits/neuron/trial: 1.39 bits per neuron per trial | Improved |
| use flexible generative model instead of Exp-Lin (Hoyer & Hyvarinen) (ablation) | NSC posterior · V1 session with 29 neurons, posterior log-likelihood | bits/neuron/trial: 0.61 bits per neuron per trial | Improved |
| use flexible generative model instead of Exp1-Lin (ablation) | NSC posterior · V1 session with 29 neurons, single-trial correlation | correlation: 10% higher | Improved |
| use flexible generative model instead of Exp-Lin (ablation) | NSC posterior · V1 session with 29 neurons, single-trial correlation | correlation: 8% higher | Improved |
| replace linear likelihood with MLP likelihood in Flow-prior model (ablation) | Flow-Lin · All 12 sessions, posterior log-likelihood | bits/neuron/trial: 0.019 bits per neuron per trial less | No clear effect |
| train discriminative system identification model on real responses instead of NSC generative model posterior | Flex-MLP posterior · V1 recordings, posterior log-likelihood | bits/neuron/trial: 0.17 bits per neuron per trial | Improved |
| train discriminative system identification model instead of NSC generative model posterior | Flex-MLP posterior · V1 recordings, single-trial correlation | correlation: 17% higher correlation | Improved |
| fit flexible model on data from classical NSC generators versus mismatched classical models | Flexible NSC model · Synthetic data (HNH, ONF, Gauss) | joint log-likelihood | Improved |
Closest papers by meaning
- Modeling Dynamic Neural Activity by combining Naturalistic Video Stimuli and Stimulus-independent Latent Factors (NeurIPS 2025 · similarity 0.89)
- Neuroformer: Multimodal and Multitask Generative Pretraining for Brain Data (ICLR 2024 · similarity 0.88)
- Decomposing stimulus-specific sensory neural information via diffusion models (NeurIPS 2025 · similarity 0.88)
- BLEND: Behavior-guided Neural Population Dynamics Modeling via Privileged Knowledge Distillation (ICLR 2025 · similarity 0.87)
- Complex priors and flexible inference in recurrent circuits with dendritic nonlinearities (ICLR 2024 · similarity 0.87)
- Benchmarking Out-of-Distribution Generalization Capabilities of DNN-based Encoding Models for the Ventral Visual Cortex. (NeurIPS 2024 · 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.
