# US 11,308,382 B2 · Neuromorphic synapses

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 11,308,382 B2
- Document ID: US11308382B2
- Status: Granted
- Kind: B2
- Application number: 15/686,870
- Filed: 2017-08-25
- Published: 2019-02-28
- Granted: 2022-04-19
- Earliest claimed: 2017-08-25
- Assignees: International Business Machines Corporation
- Applicants: International Business Machines Corporation
- Inventors: Wabe W. Koelmans; Timoleon Moraitis; Abu Sebastian
- Web page: https://patentgrep.com/patent/US11308382B2

## CPC

- G06N 3/004: Artificial life, i.e. computing arrangements simulating life
- G06N 3/049: Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
- G06N 3/063: using electronic means
- G06N 3/065: Analogue means
- G06N 3/084: Backpropagation, e.g. using gradient descent
- G06N 3/088: Non-supervised learning, e.g. competitive learning
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

Neuromorphic synapse apparatus is provided comprising a synaptic device and a control signal generator. The synaptic device comprises a memory element, disposed between first and second terminals, for conducting a signal between those terminals with an efficacy which corresponds to a synaptic weight in a read mode of operation, and a third terminal operatively coupled to the memory element. The memory element has a non-volatile characteristic, which is programmable to vary the efficacy in response to programming signals applied via the first and second terminals in a write mode of operation, and a volatile characteristic which is controllable to vary the efficacy in response to control signals applied to the third terminal. The control signal generator is responsive to input signals and is adapted to apply control signals to the third terminal in the read and write modes, in dependence on the input signals, to implement predetermined synaptic dynamics.

## Claims

### Claim 1

A neuromorphic synapse apparatus comprising:
a synaptic device comprising a memory element, disposed between first and second terminals, for conducting a signal between those terminals with an efficacy which corresponds to a synaptic weight in a read mode of operation, and a third terminal directly connected to the memory element, the memory element having a non-volatile characteristic, which is programmable to vary said efficacy in a non-volatile manner in response to programming signals applied via the first and second terminals in a write mode of operation, and a volatile characteristic which is controllable to vary said efficacy in a volatile manner in response to control signals applied to the third terminal; and
a control signal generator responsive to input signals and adapted to apply control signals to the third terminal in said read and write modes, in dependence on said input signals, to implement predetermined synaptic dynamics.

### Claim 2 (depends on 1)

The apparatus of claim 1, wherein the control signal generator is adapted to apply control signals in said read mode to implement a predetermined short-term plasticity rule via control of said volatile characteristic.

### Claim 3 (depends on 1)

The apparatus of claim 1 wherein the control signal generator is adapted to apply control signals in said write mode to implement a predetermined long-term plasticity rule via programming of said non-volatile characteristic.

### Claim 4 (depends on 1)

The apparatus of claim 1, wherein the control signal generator is adapted:
to apply control signals in said read mode to implement a predetermined short-term plasticity rule via control of said volatile characteristic; and
to apply control signals in said write mode to implement a predetermined long-term plasticity rule via programming of said non-volatile characteristic.

### Claim 5 (depends on 1)

The apparatus of claim 1, wherein the synaptic device is arranged for communication of signals between a pre-neuron circuit, operable to produce pre-neuron action signals, and a post-neuron circuit, operable to produce post-neuron action signals, and wherein said input signals comprise at least one of the pre-neuron and post-neuron action signals.

### Claim 6 (depends on 5)

The apparatus of claim 5, wherein said input signals comprise said pre-neuron action signals and wherein the control signal generator is adapted to apply said control signals in the read mode, in dependence on timing of the pre-neuron action signals, to implement a synaptic-fatigue rule via control of said volatile characteristic.

### Claim 7 (depends on 5)

The apparatus of claim 5, wherein said input signals comprise both of said pre- and post-neuron action signals and wherein the control signal generator is adapted to apply said control signals in the write mode, in dependence on relative timing of the pre- and post-neuron action signals, to implement a spike-timing dependent plasticity rule via programming of said non-volatile characteristic.

### Claim 8 (depends on 1)

The apparatus of claim 1, wherein:
the memory element comprises a resistive memory material extending between the first and second terminals and a variable-resistance channel component extending between the first and second terminals, whereby resistance of the resistive memory material can be programmed in response to said programming signals in the write mode; and
the third terminal comprises a gate terminal provided on the channel component for controlling resistance of the channel component in response to said control signals in the read and write modes.

### Claim 9 (depends on 8)

The apparatus of claim 8, wherein:
the variable-resistance channel component extends between the first and second terminals in contact with the resistive memory material, and wherein respective lengths along the channel component of high- and low-resistance regions of the resistive memory material can be varied in response to said programming signals to program resistance of the resistive memory material in the write mode; and
the control signal generator is adapted such that control signals applied to the gate terminal in the read mode provide a resistance of the channel component which is between that of said high- and low-resistance regions of the resistive memory material.

### Claim 10 (depends on 9)

The apparatus of claim 9, wherein the channel component comprises an elongate annulus, the resistive memory material is disposed within the annulus, and the gate terminal is provided on an outer surface of the annulus.

### Claim 11 (depends on 9)

The apparatus of claim 9, wherein the resistive memory material comprises a phase-change memory material, said high- and low-resistance regions comprising amorphous and crystalline phases respectively of the material.

### Claim 12 (depends on 8)

The apparatus of claim 8, wherein the control signal generator is adapted to apply control signals in the read mode to implement a predetermined short-term plasticity rule via controlling resistance of the channel component.

### Claim 13 (depends on 8)

The apparatus of claim 8, wherein the control signal generator is adapted to apply control signals in the write mode to implement a predetermined long-term plasticity rule via programming resistance of the resistive memory material.

### Claim 14 (depends on 8)

The apparatus of claim 8, wherein the control signal generator is adapted:
to apply control signals in said read mode to implement a predetermined short-term plasticity rule via controlling resistance of the channel component; and
to apply control signals in said write mode to implement a predetermined long-term plasticity rule via programming resistance of the resistive memory material.

### Claim 15 (depends on 8)

The apparatus of claim 8, wherein the synaptic device is arranged for communication of signals between a pre-neuron circuit, operable to produce pre-neuron action signals, and a post-neuron circuit, operable to produce post-neuron action signals, and wherein said input signals comprise at least one of the pre-neuron and post-neuron action signals.

### Claim 16 (depends on 15)

The apparatus of claim 15, wherein said input signals comprise said pre-neuron action signals and wherein the control signal generator is adapted to apply said control signals in the read mode, in dependence on timing of the pre-neuron action signals, to implement a synaptic-fatigue rule via controlling resistance of the channel component.

### Claim 17 (depends on 15)

The apparatus of claim 15, wherein said input signals comprise both of said pre- and post-neuron action signals and wherein the control signal generator is adapted to apply said control signals in the write mode, in dependence on relative timing of the pre- and post-neuron action signals, to implement a spike-timing dependent plasticity rule via programming resistance of the resistive memory material.