Artificial neuron and method of using same
AssigneeMotorola, Inc.
InventorShay-Ping Thomas Wang
Life of this patent
Estimating the term…
Abstract
An artificial neuron, which may be implemented either in hardware or software, has only one significant processing element in the form of a multiplier. Inputs are first fed through gating functions to produce gated inputs. These gated inputs are then multiplied together to produce a product which is multiplied by a weight to produce the neuron output.
Claims8
A neuron circuit comprising:
- a multiplier circuit in communication with a plurality of gated input signals, each of said gated input signals representing one of a plurality of inputs to said neuron circuit raised to an exponential power of one of a plurality of gating functions g.sub.i, said multiplier circuit for multiplying said gated input signals together to produce a product and for multiplying said product by a predetermined weight to generate without a threshold comparison a neuron output signal;
- wherein at least one of said plurality of gating functions g.sub.i is greater than 1.
An integrated circuit comprising:
- at least one neuron circuit comprising:
- a multiplier circuit in communication with a plurality of gated input signals, each of said gated input signals representing one of plurality of inputs to said neuron circuit raised to a power of gating function g.sub.i, wherein g.sub.i is greater than 1, said multiplier circuit for multiplying said gated input signals together to produce a product and for multiplying said product by a predetermined weight to generate without a threshold comparison a neuron output signal.
In neuron circuit having a plurality of inputs x.sub.i, wherein i is a positive integer, a method of producing a neuron output signal comprising the following steps:
- (a) receiving a first input signal from a first of said plurality of inputs;
- (b) applying a first gating function to said first input signal by raising said first input signal to a power of g.sub.i to produce a first gated input;
- (c) multiplying said first gated input by a predetermined weight to produce a product;
- (d) determining whether all of said plurality of inputs have been received;
- (i) if so, generating without a threshold comparison said neuron output signal based on the product;
- (ii) if not, proceeding to step (e);
- (e) receiving another input signal from another said plurality of inputs;
- (f) applying another gating function to said another input signal to produce another gated input;
- (g) multiplying said another gated input by said product to produce a further product;
- and
- (h) returning to step (d) until all of said plurality of inputs have been received;
- wherein at least one of said gating functions raises a corresponding input signal to a power greater than 1.
In a neuron circuit having a plurality of inputs x.sub.i, wherein i is a positive integer, a method of producing a neuron output signal, the method comprising the following steps:
- (a) applying a gating function g.sub.i to a corresponding input signal on each of said inputs x.sub.i to produce a corresponding plurality of gated inputs by raising each of said inputs x.sub.i to a power of g.sub.i ;
- (b) multiplying said gated inputs together to produce a product;
- and
- (c) generating without a threshold comparison said neuron output signal by multiplying said product by a predetermined weight;
- wherein at least one of said gating functions raises a corresponding input signal to a power greater than 1.
Dependencies are inferred from the claim text.