Method of operating a neural network

• US 5,574,827 A
• Filed: 04/17/1995
• Issued: 11/12/1996
• Est. Priority Date: 06/14/1993
• Status: Expired due to Fees

• Current Assignee: Freescale Semiconductor, Inc. (NXP Semiconductors NV)
• Original Assignee: Motorola, Inc. (Motorola Solutions, Inc.)
• Inventors: Wang, Shay-Ping T.
• Primary Examiner(s): Davis, George B.
• Application Number:
  US08/422,474
• Time in Patent Office: 575 Days
• Field of Search: 395/24, 395/23, 395/21
• US Class Current: 706/31
• CPC Class Codes: G06F 2207/5523 Calculates a power, e.g. the square, of a number or a function, e.g. polynomials
• CPC Class Codes: G06F 7/552 Powers or roots , e.g. Pythagorean sums
• CPC Class Codes: G06N 3/063 using electronic means

Abstract

A method is implemented in hardware or software type neural network, the neural network is constructed of neurons or neuron circuits each having only one significant processing element in the form of a multiplier. Each neuron applying a gating function to each of neural network inputs. The neural network utilizes a training algorithm which does not require repetitive training and which yields a global minimum to each given set of input vectors.

Claims

1. A method of operating a neural network having a plurality of neural network inputs, a plurality of neurons for inputting a plurality of inputs and generating at least one output, said method comprising the steps of:

(a) distributing each of said neural network inputs to each of said neuron inputs;

(b) each neuron applying a gating function to each of said neural network inputs to produce corresponding gated inputs;

(c) each neuron multiplying each of said gated inputs together to generate a product;

(d) each neuron multiplying said product by a weight value to generate a neuron output for each neuron; and

(e) summing the outputs of said neurons to generate said at least one network output.

2. The method of operating a neural network in claim 1 wherein the operation of said neural network is based upon a polynomial expansion.

3. The method of operating a neural network in claim 2 wherein said polynomial expansion has the form:

##EQU4## wherein y represents the output of the neural network;

wherein w_i-1 represents the weight value for the ith neuron;

wherein x₁, x₂, . . . , x_n represent inputs to said neural network;

wherein g_1i, . . . g_ni represent gating functions for the ith neuron which are applied to said inputs; and

wherein n and m are positive integers.

4. The method of operating a neural network in claim 3 wherein each x_i is represented by the function x_i =f_i (zj), wherein zj is any arbitrary variable, and wherein the indices i and j are any positive integers.

5. The method of operating a neural network in claim 3 wherein the operation of said neural network is based upon a truncated version of said polynomial expansion.

6. The method of training a neutral network in claim 3 wherein said polynomial expansion has the form:

##EQU5## wherein y represents the output of the neural network;

wherein W₀, w₁, . . . w_N-1 represents weight values, N being a positive integer representing the Nth neuron, and the term w_i in general representing the weight value associated with the i+1 neuron;

wherein x₁, x₂, . . ., x_n represent inputs to said neural network, n being a positive integer; and

wherein g_1N, . . . , g_nN represents grating functions for the Nth neuron which are applied to said inputs.

7. The method of training a neural network in claim 2 wherein said polynomial expansion has the form:

##EQU6## wherein y represents the output of the neural network;

wherein w_i-1 represents the weight value for the ith neuron;

wherein x_i, x_j, x_k represent inputs to said neural network;

wherein n is a positive integer; and

wherein f₁(i) through f₆(i,j,k) are index-generating functions.

8. The method of operating a neutral network in claim 1 wherein the operation of said network is based upon an orthogonal function.

9. The method of operating a neural network in claim 8 wherein said orthogonal function is a sine, cosine, exponential, logarithmic, Fourier transformation, Legendre polynomial, or a non-linear function such as a Volterra function or a radial basis function.

10. The method of operating a neural network in claim 1 wherein the operation of said neural network is based upon a combination of polynomial expansion and orthogonal functions.

11. The method of operating a neural network in claim 1 wherein said neural network is contained on at least one integrated circuit.

12. The method of operating a neural network in claim 1 wherein said neural network is contained in a computer program.