Machine learning with model filtering and model mixing for edge devices in a heterogeneous environment

Patent metadata

Field	Value
Publication number	US-10387794-B2
Application number	US-201514602843-A
Country	US
Kind code	B2
Filing date	Jan 22, 2015
Priority date	Jan 22, 2015
Publication date	Aug 20, 2019
Grant date	Aug 20, 2019

Abstract

Machine learning with model filtering and model mixing for edge devices in a heterogeneous environment is disclosed. In an example embodiment, an edge device includes a communication module, a data collection device, a memory, a machine learning module, and a model mixing module. The edge device analyzes collected data with a model for a first task, outputs a result, and updates the model to create a local model. The edge device communicates with other edge devices in a heterogeneous group, transmits a request for local models to the heterogeneous group, and receives local models from the heterogeneous group. The edge device filters the local models by structure metadata, including second local models, which relate to a second task. The edge device performs a mix operation of the second local models to generate a mixed model which relates to the second task, and transmits the mixed model to the heterogeneous group.

First claim

The invention is claimed as follows:

1. An edge device comprising:
a communicator configured to communicate with a plurality of edge devices;
a data collector configured to collect data;
a memory configured to store the data collected by the data collector; and
one or more processors;
wherein the edge device is configured to:
analyze, by the one or more processors, using a local model, the data collected by the data collector;
transmit, by the communicator, requests for local models to the plurality of edge devices;
receive, by the communicator, a first plurality of local models from the plurality of edge devices, each of the first plurality of local models being updated by each of the plurality of edge devices based on analysis result of data corrected by the each of the plurality of edge devices;
filter, by the one or more processors, the first plurality of local models by at least one of structure metadata, context metadata, and data distribution;
select, by the one or more processors, a second plurality of local models from the first plurality of local models based on a result of the filtering;
generate, by the one or more processors, a mixed model from the second plurality of local models; and
transmit, by the communicator, the mixed model to other edge devices.

2. The edge device of claim 1 , wherein the mix operation is determined using a size of a group of the second plurality of local models and a computer power of the edge device.

3. The edge device of claim 1 , wherein the mix operation is an averaging operation.

4. The edge device of claim 1 , wherein the mix operation is a genetic algorithm operation.

5. The edge device of claim 1 , wherein the mix operation is an enumeration operation.

6. The edge device of claim 1 , wherein the mix operation is an ensemble operation.

7. The edge device of claim 1 , wherein the edge device is further configured to replace the local model of the edge device with the mixed model.

8. The edge device of claim 1 , wherein the edge device is one of a shopping cart device and a surveillance camera.

9. The edge device of claim 1 , wherein the edge device is incorporated in an automobile.

10. The edge device of claim 1 , wherein the edge device is an automatic teller machine.

11. The edge device of claim 1 , wherein the plurality of edge devices comprise:
a first edge device analyzing, using a local model of the first edge device, data collected by the first edge device; and
a second edge device analyzing, using a local model of the second edge device, data collected by the second edge device, wherein the data collected by the second edge device are different from the data collected by the first edge device.

12. The edge device of claim 1 , wherein the edge device updates the local model of the edge device based on analysis result by the edge device itself.

13. The edge device of claim 1 , wherein the first plurality of local models are update from a global model different from any of the first plurality of local models.

14. A method comprising:
transmitting, by a communicator in an edge device, requests for local models to a plurality of edge devices;
receiving, by the communicator, a first plurality of local models from the plurality of edge devices, each of the first plurality of local models being updated by each of the plurality of edge devices based on analysis result of data collected by the each of the plurality of edge devices;
filtering, by one or more processors in the edge device, the first plurality of local models by at least one of structure metadata, context metadata, and data distribution;
selecting, by the one or more processors, a second plurality of local models from the first plurality of local models based on a result of the filtering,
generating, by the one or more processors, a mixed model from the second plurality of local models; and
transmitting, by the communicator, the mixed model to other edge devices.

15. An edge device comprising:
a memory; and
one or more processors coupled to the memory and configured to:
transmit requests for local models to other edge devices,
receive first local models from the other edge devices, each of the first local models being updated by each of the other edge devices based on analysis result of data collected by the each of the other edge devices,
select second local models by filtering the first local models based on at least one of structure metadata, context metadata, and data distribution,
generate a mixed model from the second local models, and
transmit the mixed model to the other edge devices.

16. The edge device of claim 15 , wherein the other edge devices comprises:
a first edge device analyzing, using a local model of the first edge device, data collected by the first edge device; and
a second edge device analyzing, using a local model of the second edge device, data collected by the second edge device, wherein the data collected by the second edge device are different from the data collected by the first edge device.

17. The edge device of claim 15 , wherein the edge device updates the local model of the edge device based on analysis result by the edge device itself.

18. The edge device of claim 15 , wherein the first local models are updated from a global model different from the first local models.

19. The edge device of claim 15 , wherein the each of the first local models being updated by the each of the other edge devices based on a correctness of the analysis result generated by the each of the other edge devices.

20. The edge device of claim 15 , wherein the one or more processors further configured to:
collect data and store the data in the memory, and
analyze, using a local model, the data collected.

Assignees

• Preferred Networks Inc

Inventors

• Okanohara Daisuke
• Clayton Justin B
• Nishikawa Toru
• Hido Shohei
• Kubota Nobuyuki
• Ota Nobuyuki
• Tokui Seiya

Classifications

• G06N20/00 Primary

  Machine learning · CPC title