Spike-Based Population Coding and Working Memory
Author Information
Author(s): Martin Boerlin, Sophie Denève
Primary Institution: École Normale Supérieure, Paris, France
Hypothesis
How can populations of spiking neurons implement probabilistic computations for optimal decision-making?
Conclusion
The study proposes a model where spiking neurons can optimally combine sensory information and maintain working memory by representing entire probability distributions.
Supporting Evidence
- Neurons can represent and compute with probability distributions.
- The model shows that sustained activity can maintain information over time.
- Spike times signal a prediction error, allowing for optimal sensory integration.
- Population responses can represent entire probability distributions, not just single values.
Takeaway
This study shows that groups of brain cells can work together to remember things and make decisions even when things are uncertain.
Methodology
The authors developed a model using recurrent networks of integrate-and-fire neurons to simulate sensory integration and working memory.
Limitations
The model assumes specific dynamics for the stimulus and may not account for all real-world complexities.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website