Research map / Architectures and efficiency / CNN design
Learning long-range spatial dependencies with horizontal gated recurrent units
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 one hGRU layer with horizontal recurrent interactions | Pathfinder models · Pathfinder, all tested path lengths | accuracy | Helped |
| Use an hGRU instead of feedforward hierarchical models | Pathfinder models · Pathfinder, 14-paddle paths | area under the learning curve (ALC) | Helped |
| Use an hGRU instead of deeper feedforward models | Pathfinder models · Pathfinder, 14-paddle paths | parameter efficiency: 200× | Helped |
| Increase hGRU processing time (ablation) | hGRU · Pathfinder, longer paths | accuracy | Helped |
| Add linear and quadratic horizontal interactions (ablation) | hGRU · Pathfinder, all tested path lengths | accuracy | Helped |
| Add batch normalization (ablation) | hGRU · Pathfinder | accuracy | Helped |
| Replace the squashing nonlinearity with ReLU and use separate excitation and inhibition kernels (ablation) | hGRU · Pathfinder | accuracy | Helped |
| Use a two-layer convolutional GRU candidate-state computation (ablation) | GRU · Pathfinder, 6-paddle paths | accuracy | Helped |
| Use a two-layer convolutional GRU candidate-state computation (ablation) | GRU · Pathfinder, 9- and 14-paddle paths | accuracy | Hurt |
| Add dilated or non-local operations to feedforward models (ablation) | Feedforward models · Pathfinder | accuracy | No clear effect |
| Increase feedforward model depth (ablation) | Feedforward models · Pathfinder, longer paths | accuracy | Mixed |
| Increase path length | All tested models · Pathfinder | accuracy | Hurt |
Closest papers by meaning
- Task-Driven Convolutional Recurrent Models of the Visual System (NeurIPS 2018 · similarity 0.87)
- Hierarchically Gated Recurrent Neural Network for Sequence Modeling (NeurIPS 2023 · similarity 0.87)
- Stable and expressive recurrent vision models (NeurIPS 2020 · similarity 0.86)
- Unveiling the Spatial-temporal Effective Receptive Fields of Spiking Neural Networks (NeurIPS 2025 · similarity 0.85)
- ViG: Linear-complexity Visual Sequence Learning with Gated Linear Attention (AAAI 2025 · similarity 0.84)
- Fully Hyperbolic Convolutional Neural Networks for Computer Vision (ICLR 2024 · similarity 0.84)
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.