Related Experiment Videos
An algebraic interpretation of PSP composition
1Supélec, Signal Processing and Neural Networks Team, Cesson-Sévigné, France. Gilles.Vaucher@supelec.fr
Bio Systems
|January 14, 1999
Summary
This study introduces a novel artificial neuron model incorporating time and biological properties. The new model treats postsynaptic potentials as two-degree-of-freedom elements, enabling asynchronous computation and potentially modeling biological neuron electrical properties.
Area of Science:
- Computational Neuroscience
- Artificial Intelligence
- Biophysics
Background:
- Integrating temporal dynamics into artificial neurons (ANs) is a significant challenge.
- Existing AN models, like McCulloch and Pitts, lack inherent temporal processing.
- Biological neurons exhibit complex temporal behaviors crucial for neural computation.
Purpose of the Study:
- To develop a novel artificial neuron model that incorporates temporal dynamics.
- To bridge the gap between biological neural properties and algebraic AN models.
- To propose a framework for asynchronous computation using interconnected neurons.
Main Methods:
- Combining properties of biological neuron models with McCulloch and Pitts AN algebra.
- Representing postsynaptic potentials (PSPs) as numerical elements with two degrees of freedom.
- Developing an algebra of impulses to formalize the functioning of the proposed neuron network.
Main Results:
- A new extended artificial neuron model that integrates temporal aspects.
- Demonstration of how a network of these neurons can function as an asynchronous computer.
- Formalization of neural network operations through an algebra of impulses.
Conclusions:
- The proposed model offers a new perspective on temporal dynamics in ANs.
- This approach could facilitate the development of asynchronous computing architectures.
- The model shows potential for simulating passive electrical properties of biological neurons.