18264427. Neuromorphic circuit and associated training method simplified abstract (CENTRE NATIONAL DE LA RECHERCHE SCIENTIFIQUE)

From WikiPatents
Jump to navigation Jump to search

Neuromorphic circuit and associated training method

Organization Name

CENTRE NATIONAL DE LA RECHERCHE SCIENTIFIQUE

Inventor(s)

Julie Grollier of GENTILLY (FR)

Erwann Martin of PALAISEAU CEDEX (FR)

Damien Querlioz of ARCUEIL (FR)

Teodora Petrisor of PALAISEAU CEDEX (FR)

Neuromorphic circuit and associated training method - A simplified explanation of the abstract

This abstract first appeared for US patent application 18264427 titled 'Neuromorphic circuit and associated training method

Simplified Explanation: The patent application describes a neuromorphic circuit that implements a spiking neural network with bidirectional synapses made of memristors. Neurons fire spikes at varying rates and are connected via synapses. The circuit includes a training module that estimates the time derivative of spike rates, adjusts the position of the interconnection between synapse and neuron, and sends control signals to optimize connections.

Key Features and Innovation:

  • Implementation of a spiking neural network with bidirectional synapses using memristors.
  • Neurons firing spikes at variable rates and connected via synapses.
  • Training module estimating time derivative of spike rates and adjusting interconnections.
  • Control signals sent to optimize connections between neurons and synapses.

Potential Applications: This technology could be used in:

  • Artificial intelligence systems
  • Robotics
  • Brain-computer interfaces
  • Neuromorphic computing

Problems Solved:

  • Optimization of connections in spiking neural networks.
  • Efficient training of neural networks.
  • Real-time adjustment of synapse connections.

Benefits:

  • Improved performance of neural networks.
  • Enhanced adaptability and learning capabilities.
  • Energy-efficient computing.

Commercial Applications: Potential commercial uses include:

  • Neuromorphic processors for AI applications
  • Neuromorphic hardware for robotics
  • Neuromorphic chips for brain-computer interfaces

Questions about Neuromorphic Circuits: 1. How do neuromorphic circuits differ from traditional computing systems? 

  - Neuromorphic circuits mimic the structure and function of the human brain, enabling parallel processing and efficient learning.

2. What are the advantages of using memristors in implementing neural networks? 

  - Memristors offer non-volatile memory and can mimic synaptic plasticity, making them ideal for neuromorphic applications.


Original Abstract Submitted

A neuromorphic circuit implementing a spiking neural network and including bidirectional synapses made by a set of memristors arranged in an array, neurons firing spikes at a variable rate and connected to neurons via a synapse, and a neural network training module including, for at least one bidirectional synapse, an estimation unit obtaining an estimation of the time derivative of the spike rate of each neuron, an interconnection having at least two positions between the synapse and each neuron, and a controller sending a control signal to the interconnection after a spike, the signal changing the position of the interconnection, so as to connect the estimation unit and the synapse.

Retrieved from "http://wikipatents.org/index.php?title=18264427._Neuromorphic_circuit_and_associated_training_method_simplified_abstract_(CENTRE_NATIONAL_DE_LA_RECHERCHE_SCIENTIFIQUE)&oldid=233137"