Related Experiment Videos
Reading population codes: a neural implementation of ideal observers.
S Deneve1, P E Latham, A Pouget
1Brain and Cognitive Science Department, University of Rochester, Rochester, New York 14627, USA.
Nature Neuroscience
|July 21, 1999
Summary
Researchers show that the brain can approximate maximum likelihood (ML) decoding of neural population codes, even with noisy neuronal responses. This suggests cortical circuits may act as ideal observers, efficiently extracting sensory information.
Area of Science:
- Neuroscience
- Computational Neuroscience
Background:
- Sensory and motor information is encoded by neuronal populations with bell-shaped tuning curves.
- Extracting this information is challenging due to noise in neuronal responses.
Purpose of the Study:
- To investigate if a biologically plausible model can approximate maximum likelihood (ML) decoding.
- To determine if cortical circuitry can function as an ideal observer.
Main Methods:
- Simulations and analysis of neuronal population activity.
- Modeling of cortical circuitry with nonlinear activation functions.
Main Results:
- A biologically plausible model closely approximates ML decoding.
- The approximation holds across various nonlinear activation functions.
Conclusions:
- Cortical areas may generally function as ideal observers.
- Efficient information extraction from noisy neural population codes is achievable.