×

Method for supervised teaching of a recurrent artificial neural network

  • US 20040015459A1
  • Filed: 05/29/2003
  • Published: 01/22/2004
  • Est. Priority Date: 10/13/2000
  • Status: Active Grant
First Claim
Patent Images

1. A method for constructing a discrete-time recurrent neural network and training it in order to minimize its output error, comprising the steps a. constructing a recurrent neural network as a reservoir for excitable dynamics (DR network);

  • b. providing means of feeding input to the DR network;

    c. attaching output units to the DR network through weighted connections;

    d. training the weights of the connections from the DR network to the output units in a supervised training scheme.

View all claims
  • 1 Assignment
Timeline View
Assignment View
    ×
    ×