Battery-based neural network weights

Patent metadata

Field	Value
Publication number	US-11023802-B2
Application number	US-201715803957-A
Country	US
Kind code	B2
Filing date	Nov 6, 2017
Priority date	Mar 13, 2017
Publication date	Jun 1, 2021
Grant date	Jun 1, 2021

Abstract

Methods for controlling the resistance of a controllable resistive element include determining an amount of electrical resistance change for the controllable resistive element. A concentration difference is determined for a charge carrier ion in a resistor layer of the controllable resistance element that corresponds to the electrical resistance change for the controllable resistive element. A duration and amplitude of a current pulse is determined that changes the charge carrier ion concentration by the determined difference. A positive or negative current pulse is applied to a controllable resistive element for the determined duration.

First claim

The invention claimed is:

1. A method for adjusting the electrical resistance of a controllable resistive element, comprising:
determining an amount of electrical resistance change for the controllable resistive element;
determining a concentration difference for a charge carrier ion in a resistor layer of the controllable resistive element that corresponds to the electrical resistance change for the controllable resistive element;
determining a duration and amplitude of a current pulse that changes the charge carrier ion concentration by the determined difference;
applying a positive or negative current pulse to a controllable resistive element for the determined duration, thereby causing charge carrier ions to travel through an electrolyte layer of the controllable resistive element, to or from the resistor layer; and
waiting a predetermined time before applying a read voltage to the controllable resistive element to allow charge carrier ions in the resistor layer to settle.

2. The method of claim 1 , wherein applying the positive or negative current pulse comprises applying a current pulse between a write electrode of the controllable resistive element and the shared read/write electrode of the controllable resistive element to decrease the charge carrier ion concentration by the determined difference.

3. The method of claim 1 , wherein applying the positive or negative current pulse comprises applying a current pulse between a write electrode of the controllable resistive element and the shared read/write electrode of the controllable resistive element to increase the charge carrier ion concentration by the determined difference.

4. The method of claim 1 , wherein applying the positive or negative pulse causes charge carrier ions to move from or to a reservoir layer.

5. The method of claim 1 , wherein the resistor layer is formed from a lithium-containing material.

6. The method of claim 5 , wherein the lithium-containing material is selected from the group consisting of LiCoO 2 , LiNbO 3 , LiMnO 2 , LiV 2 O 5 , LiFePO 4 , LiNi x Mn y Co z , V 2 O 5 —LiBO 2 , Li 4 Ti 5 O 12 , Li x Al, Li x C, and Li x Si.

7. The method of claim 1 , wherein the electrolyte layer is formed from a material selected from the group consisting of lithium phosphorous oxy-nitride and an organic material-based electrolyte.

8. The method of claim 1 , wherein applying the positive or negative current pulse to the controllable resistive element causes charge carrier ions to travel from or to a reservoir layer.

9. The method of claim 8 , wherein the reservoir layer is formed from a lithium compound material selected from the group consisting of LiCoO 2 , LiNbO 3 , LiMnO 2 , LiV 2 O 5 , LiFePO 4 , LiNi x Mn y Co z , V 2 O 5 —LiBO 2 , Li 4 Ti 5 O 12 , Li 2 TiO 3 , Li x Al, Li x C, Li, and Li x Si.

Assignees

• IBM

Inventors

• De Souza Joel P
• Lee Yun Seog
• Li Ning
• Sadana Devendra K

Classifications

• G06N3/065

  Analogue means · CPC title

• G06N3/063 Primary

  using electronic means · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06N3/0499

  Feedforward networks · CPC title

• G06N3/084 Primary

  Backpropagation, e.g. using gradient descent · CPC title