Neural network processing apparatus for identifying an unknown image pattern as one of a plurality of instruction image patterns
First Claim
Patent Images
1. A neural network processing apparatus, comprising:
- a plurality of neurons functioning as a group to generate an output signal by performing an operating function on an input signal, the operating function having a gain;
a network for interconnecting the plurality of neurons and for setting weights according to the strengths of connections between the plurality of neurons;
activator means for activating each neuron of the plurality of neurons to receive as a neuron input signal the weighted sum of the output signals of interconnected other neurons of the plurality of neurons and corresponding weight between the plurality of neurons, and to generate the output signals;
calculation means for repeatedly calculating an average of the absolute values of differences between the output values of all of the plurality of neurons and a center value, and for calculating a difference between a current value of the average and a previous value of the average;
function control means for determining the gain of the operating function according to the difference between the current value of the average and a threshold, and according to the difference between the current value of the average and the previous value of the average; and
function set means for setting the gain of each neuron using the gain of the operating function determined by the function control means for each neuron.
1 Assignment
0 Petitions
Accused Products
Abstract
A neural network processing apparatus calculates an average of the absolute values of differences between the output values of all neurons and a center value whenever the output value of all neurons change, and calculates the difference between the average and the previous average. If the average is larger than a threshold or the previous average, the gain of a function in the network is decreased. If the average is smaller than the threshold or the previous average, the gain of the function is increased. Then the controlled function is set to each neuron and the neural network is activated repeatedly to correctly identify an unknown multivalued image pattern.
-
Citations
17 Claims
-
1. A neural network processing apparatus, comprising:
-
a plurality of neurons functioning as a group to generate an output signal by performing an operating function on an input signal, the operating function having a gain; a network for interconnecting the plurality of neurons and for setting weights according to the strengths of connections between the plurality of neurons; activator means for activating each neuron of the plurality of neurons to receive as a neuron input signal the weighted sum of the output signals of interconnected other neurons of the plurality of neurons and corresponding weight between the plurality of neurons, and to generate the output signals; calculation means for repeatedly calculating an average of the absolute values of differences between the output values of all of the plurality of neurons and a center value, and for calculating a difference between a current value of the average and a previous value of the average; function control means for determining the gain of the operating function according to the difference between the current value of the average and a threshold, and according to the difference between the current value of the average and the previous value of the average; and function set means for setting the gain of each neuron using the gain of the operating function determined by the function control means for each neuron. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16)
-
-
17. A neural network processing apparatus, comprising:
-
a plurality of neurons functioning as a group to generate an output signal by performing an operating function on an input signal, the operating function having a gain; a network for interconnecting the plurality of neurons and for setting weights according to the strengths of connections between the plurality of neurons; activator means for activating each neuron of the plurality of neurons to receive as a neuron input signal the weighted sum of the output signals of interconnected other neurons of the plurality of neurons and corresponding weight between the plurality of neurons, and to generate the output signals; calculation means for calculating an average of the absolute values of differences between the output values of all of the plurality of neurons and a center value whenever the output values of all of the plurality of neurons change, and for calculating a difference between a current value of the average and a previous value of the average; function control means for determining the gain of the operating function by decreasing the gain of the operating function when the current value of the average is larger than a threshold or the previous value of the average, and by increasing the gain of the operating function when the current value of the average is smaller than the threshold or the previous value of the average; and function set means for setting the gain of each neuron using the gain of the operating function determined by the function control means for each neuron.
-
Specification