Separate Learning System and Method Using Two-Layered Neural Network Having Target Values for Hidden Nodes

Abstract

Disclosed herein is a separate learning system and method using a two-layered neural network having target values for hidden nodes. The separate learning system of the present invention includes an input layer for receiving training data from a user, and including at least one input node. A hidden layer includes at least one hidden node. A first connection weight unit connects the input layer to the hidden layer, and changes a weight between the input node and the hidden node. An output layer outputs training data that has been completely learned. The second connection weight unit connects the hidden layer to the output layer, changing a weight between the output and the hidden node, and calculates a target value for the hidden node, based on a current error for the output node. A control unit stops learning, fixes the second connection weight unit, turns a learning direction to the first connection weight unit, and causes learning to be repeatedly performed between the input node and the hidden node if a learning speed decreases or a cost function increases due to local minima or plateaus when the first connection weight unit is fixed and learning is performed using only the second connection weight unit, thus allowing learning to be repeatedly performed until learning converges to the target value for the hidden node.

1 . (canceled) 
   
   
       2 . (canceled) 
   
   
       3 . (canceled) 
   
   
       4 . (canceled) 
   
   
       5 . (canceled) 
   
   
       6 . (canceled) 
   
   
       7 . A separate learning method using a two-layered neural network having target values for hidden nodes, comprising the steps of:
 (a) performing learning in a second connection weight unit using training data;   (b) determining whether learning has converged when a learning speed decreases due to local minima and plateaus, and stopping the learning if it is determined that learning has converged, otherwise turn a learning direction to a first connection weight unit and allowing learning to be performed between all of the input node at least one hidden node;   (c) determining whether learning has reached a target value for the hidden node set by the first connection weight unit;   (d) turning a learning direction to the second connection weight unit and performing learning between the hidden node and at least one output node if it is determined that learning has not reached the target value for the hidden node as a result of the determination; and   (e) causing learning, performed in the second connection weight unit, to reach a global minimum.   
   
   
       8 . The separate learning method according to  claim 7 , further comprising the step of (a- 1 ) receiving training data through the input layer to train a neural network before step (a). 
   
   
       9 . The separate learning method according to  claim 7 , wherein step (b) comprises the steps of:
 (b- 1 ) selecting an output node having a largest error value with respect to the hidden node if it is determined that learning has not converged;   (b- 2 ) calculating the target value for the hidden node so that learning can reach a global minimum; and
 (b- 3 ) transmitting the error value for the hidden node and the target value for the hidden node to the first connection weight unit.