Programmable neural circuit implementable in CMOS very large scale integration

Claim 2

A neural network circuit comprising: a plurality of neuron circuits, each neuron circuit having an input node for receiving an input signal, an output node for generating an output signal and a self-feedback control node for receiving a self-feedback signal, said output signal of each neuron circuit being a function of said input signal and said self-feedback signal; and a plurality of interconnection devices, one interconnection device connected between said input nodes of each pair of neuron circuits, each interconnection device having an electrically controllable nonlinear conductance.

Claim 3

The neural network circuit claimed in claim 1, wherein: each of said interconnection devices consists of a field effect transistor having a drain connected to the input node of a first of said neuron circuits, a source connected to the input node of a second of said neuron circuits, a gate and a substrate connected to a reference voltage, said nonlinear conductance being electrically controllable via the voltage on said gate.

Claim 4

The neural network circuit claimed in claim 1, wherein: each of said plurality of neuron circuits includes a capacitor connected between said input node and said reference voltage, a resistor connected between said input node and said reference voltage, an amplifier circuit having an input terminal connected to said input node and an output terminal connected to said output node, said amplifier generating an output signal on said output terminal which is a function of the signal on said input terminal, and a field effect feedback device having a source-drain path connected between the input terminal of said amplifier circuit and the output terminal of said amplifier circuit, a gate connected to said self-feedback control node, and a substrate connected to said reference voltage, said field effect feedback device for providing variable conductance feedback between said output terminal and said input terminal corresponding to said self-feedback signal.

Claim 5

The neural network circuit claimed in claim 3, wherein: said amplifier circuit of each of said plurality of neuron circuits consists of two field effect transistor inverter circuits disposed in tandem.

Claim 6

The neural network circuit claimed in claim 3, wherein: said neural network circuit is constructed in complementary metal oxide semiconductor integrated circuits.

Claim 7

The neural network circuit claimed in claim 5, wherein: said capacitor of each of said neuron circuits consists of the parasitic capacitance of the corresponding input node.

Claim 8

The neural network circuit claimed in claim 5, wherein: said resistor of each of said neuron circuits consists of the parasitic resistance to ground of the corresponding input node.

Claim 9

A neural network circuit comprising: a plurality of neuron circuits, each neuron circuit including an input node for receiving an input signal, an output node, a self-feedback control node for receiving a self-feedback signal, a capacitor connected between said input node and a reference voltage, a resistor connected between said input node and said reference voltage, an amplifier circuit having an input terminal connected to said input node and an output terminal connected to said output node, said amplifier generating an output signal on said output terminal which is a function of said input signal on said input terminal, a field effect feedback device having a source-drain path connected between the input terminal of said amplifier circuit and the output terminal of said amplifier circuit, a gate connected to said self-feedback control node, and a substrate connected to said reference voltage, said field effect feedback device for providing variable conductance feedback between said output terminal and said input terminal corresponding to said self-feedback signal; and a plurality of interconnection field effect transistors, one interconnection field effect transistor connected between said input nodes of each pair of neuron circuits, each interconnection field effect transistor having a drain connected to the input node of a first of said neuron circuits, a source connected to the input node of a second of said neuron circuits, a gate and a substrate connected to said reference voltage.

Claim 10

A neural network circuit comprising: a plurality of neuron circuits, each neuron circuit including an input node for receiving an input signal, an output node, a self-feedback control node for receiving a self-feedback signal, a capacitor connected between said input node and a reference voltage, a resistor connected between said input node and said reference voltage, an amplifier circuit having an input terminal connected to said input node and an output terminal connected to said output node, said amplifier generating an output signal on said output terminal which is a function of said input signal on said input terminal, a field effect feedback device having a source-drain path connected between the input terminal of said amplifier circuit and the output terminal of said amplifier circuit, a gate connected to said self-feedback control node, and a substrate connected to said reference voltage, said field effect feedback device for providing variable conductance feedback between said output terminal and said input terminal corresponding to said self-feedback signal; and a plurality of interconnection field effect transistors, one interconnection field effect transistor connected between said input nodes of each pair of neuron circuits, each interconnection field effect transistor having a drain connected to the input node of a first of said neuron circuits, a source connected to the input node of a second of said neuron circuits, a gate and a substrate connected to said reference voltage.

Claim 11

The neural network circuit claimed in claim 8, wherein: said amplifier circuit of each of said plurality of neuron circuits consists of two field effect transistor inverter circuits disposed in tandem.

Claim 12

The neural network circuit claimed in claim 8, wherein: said neural network circuit is constructed in complementary metal oxide semiconductor integrated circuits.

Claim 13

The neural network circuit claimed in claim 10, wherein: said capacitor of each of said neuron circuits consists of the parasitic capacitance of the corresponding input node.

Claim 14

The neural network circuit claimed in claim 10, wherein: said resistor of each of said neuron circuits consists of the parasitic resistance to ground of the corresponding input node.