Programmable artificial neuron and associated programming method

Patent metadata

Field	Value
Publication number	US-10861545-B2
Application number	US-201916527974-A
Country	US
Kind code	B2
Filing date	Jul 31, 2019
Priority date	Aug 1, 2018
Publication date	Dec 8, 2020
Grant date	Dec 8, 2020

Abstract

A programmable artificial neuron emitting an output signal controlled by at least one control parameter, includes, for each control parameter, a capacitor and at least one block including at least one multiplexer configured to be in two states: a programming state and an operating state; a transistor; and a non-volatile resistive random access memory connected in series with the transistor, the capacitor and the resistive random access memory being mounted in parallel. The multiplexer is configured to, when it is in the programming state, set a resistance value of the resistive random access memory to set the value of the control parameter; when it is in the operating state, conserve the set resistance value of the resistive random access memory.

First claim

The invention claimed is:

1. Programmable artificial neuron emitting an output signal controlled by at least one control parameter, comprising, for each control parameter of the at least one control parameter, a capacitor and at least one block associated with said control parameter of the at least one control parameter, said at least one block including:
at least one multiplexer configured to be in two states: a programming state of said at least one block and an operating state;
a transistor; and
a non-volatile resistive random access memory connected in series with said transistor, said capacitor and said non-volatile resistive random access memory being mounted in parallel; said at least one multiplexer being configured to:
when it is in the programming state, electrically disconnect the at least one block from the programmable artificial neuron, electrically connect the gate of the transistor of the at least one block to a voltage source and electrically connect the resistive random access memory of the at least one block to a programming source;
when it is in the operating state, electrically connect the at least one block to the programmable artificial neuron.

2. The programmable artificial neuron according to claim 1 , wherein the transistor is a MOS transistor.

3. The programmable artificial neuron according to claim 1 , wherein the non-volatile resistive random access memory is an OxRAM, CBRAM or PCRAM type memory.

4. The programmable artificial neuron according to claim 1 , wherein, when the control parameter of the at least one control parameter is a time constant, the programmable artificial neuron comprises one block and one capacitor.

5. The programmable artificial neuron according to claim 1 , wherein, when the control parameter of the at least one control parameter is a refractory period, the programmable artificial neuron comprises one block and one capacitor.

6. The programmable artificial neuron according to claim 1 , wherein, when the control parameter of the at least one control parameter is a spike frequency adaptation behaviour, the programmable artificial neuron comprises two blocks and one capacitor.

7. A method for programming a programmable artificial neuron according to claim 1 , comprising for each block of the at least one block, the following steps:
electrically disconnecting the block from the programmable artificial neuron, electrically connecting the gate of the transistor of the block to a voltage source and electrically connecting the resistive random access memory of the block to a programming source by placing each multiplexer of the at least one multiplexer of the block in a programming state;
applying a voltage to the gate of the transistor of the block with the voltage source to set the value of the drain-source resistance of the transistor;
applying a programming voltage or a current with the programming source to adjust the resistance value of the non-volatile resistive random access memory of the block, and
electrically connecting the block to the programmable artificial neuron by placing each multiplexer of the at least one multiplexer in an operating state.

Assignees

• Commissariat Energie Atomique

Inventors

• Dalgaty Thomas
• Vianello Elisa

Classifications

• G06N3/049 Primary

  Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs · CPC title

• G06N3/065

  Analogue means · CPC title

• G11C13/0007

  comprising metal oxide memory material, e.g. perovskites · CPC title

• G11C13/0061

  Timing circuits or methods · CPC title

• G11C2013/0071

  Write using write potential applied to access device gate · CPC title