A Theory of Neural Computation Based on Prediction by Neurons
Author Information
Author(s): Christopher D. Fiorillo
Primary Institution: Stanford University
Hypothesis
The computational goal of the nervous system is to minimize uncertainty about future reward.
Conclusion
The theory suggests that neurons operate under shared computational principles to predict aspects of the world related to future rewards.
Supporting Evidence
- The theory integrates principles from Bayesian probability and reinforcement learning.
- Neurons are proposed to select inputs based on their correlation with future rewards.
- Prediction errors are used to drive learning and plasticity in neurons.
Takeaway
This study explains how neurons learn to predict things that will help them get rewards, like food, by using information from their environment.
Methodology
The study outlines a theoretical framework combining Bayesian probability theory with biophysical properties of neurons.
Limitations
The theory may not apply universally to all types of neurons and does not account for all factors influencing neuronal function.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website