RealFeedApp Network processes angular and linear velocity separately
Researchers developed a hierarchical recurrent neural network that processes angular velocity and linear speed separately. The model reproduces properties of all key navigation cell types, including head direction cells and grid cells, while demonstrating attractor dynamics.
- Field
- the brain and neuroscience
- What they did
- Researchers developed a hierarchical recurrent neural network that learns path integration by separately processing angular velocity and linear speed. This model reproduces the activity of key brain cell types, such as head direction cells and grid cells, while recovering their characteristic connectivity patterns.
- Why it matters
- This helps better understand how the brain self-locates in space and offers new ideas for creating neural networks capable of autonomous navigation.
#path integration#grid cells#recurrent neural network#neural coding#navigation
