Research map / Architectures and efficiency / RNNs and normalization
A mechanistic multi-area recurrent network model of decision-making
Michael Kleinman, Chandramouli Chandrasekaran, Jonathan Kao
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 3 areas instead of a single-area RNN (ablation) | mechanistic multi-area RNN · Checkerboard Task | PMd resemblance of last area | Improved |
| add Dale's law with anatomically-informed feedforward inhibition (ablation) | mechanistic multi-area RNN · Checkerboard Task | color decode accuracy in Area 3 | Improved |
| reduce feedforward E-to-I connections to 2% or less (ablation) | mechanistic multi-area RNN · Checkerboard Task | color decode accuracy in Area 3: 2% | Improved |
| lower feedforward connections to 1% without Dale's law (ablation) | mechanistic multi-area RNN · Checkerboard Task | color decode accuracy in Area 3: 1% | No clear effect |
| vary feedforward E-E and feedback connection percentages (ablation) | mechanistic multi-area RNN · Checkerboard Task | color decode accuracy | No clear effect |
| use at least 3 areas (ablation) | mechanistic multi-area RNN · Checkerboard Task | PMd-like last area | Improved |
| vary ML hyperparameters (units, learning rate) (ablation) | mechanistic multi-area RNN · Checkerboard Task | minimal color representation | No clear effect |
| train without Dale's law or with 2 areas (ablation) | mechanistic multi-area RNN · Checkerboard Task | direction axis alignment with top singular vectors | Hurt |
| double 10% of inhibitory weights to the right pool | mechanistic multi-area RNN · Area 3 psychometric curve | choice bias: 10% | Improved |
| later areas (Area 1 to Area 3) | mechanistic multi-area RNN · Checkerboard Task | color decode accuracy: 0.93, 0.76, 0.51 | Improved |
Closest papers by meaning
- Scaling Laws and Tradeoffs in Recurrent Networks of Expressive Neurons (NeurIPS 2026 · similarity 0.87)
- Emergent mechanisms for long timescales depend on training curriculum and affect performance in memory tasks (ICLR 2024 · similarity 0.87)
- A Multi-Region Brain Model to Elucidate the Role of Hippocampus in Spatially Embedded Decision-Making (ICML 2025 · similarity 0.87)
- Geometry of naturalistic object representations in recurrent neural network models of working memory (NeurIPS 2024 · similarity 0.86)
- Brain-Like Processing Pathways Form in Models With Heterogeneous Experts (NeurIPS 2025 · similarity 0.85)
- Flexible Context-Driven Sensory Processing in Dynamical Vision Models (NeurIPS 2024 · 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.
