Systems and methods for determining an artificial intelligence model in a communication system

Patent metadata

Field	Value
Publication number	US-11334801-B2
Application number	US-201816189936-A
Country	US
Kind code	B2
Filing date	Nov 13, 2018
Priority date	Nov 13, 2018
Publication date	May 17, 2022
Grant date	May 17, 2022

Abstract

A device for obtaining a local optimal AI model may include an artificial intelligence (AI) chip and a processing device configured to receive a first initial AI model from the host device. The device may load the initial AI model into the AI chip to determine a performance value of the AI model based on a dataset, and determine a probability that a current AI model should be replaced by the initial AI model. The device may determine, based on the probability, whether to replace the current AI model with the initial AI model. If it is determined that the current AI model be replaced, the device may replace the current AI model with the initial AI model. The device may repeat the above processes and obtain a final current AI model. The device may transmit the final current AI model to the host device.

First claim

We claim:

1. A device comprising:
an artificial intelligence (AI) chip;
at least one processor; and
a non-transitory computer readable medium containing programming instructions that, when executed, will cause the at least one processor to:
(i) access a dataset;
(ii) receive an initial artificial intelligence (AI) model from a host device;
(iii) load the initial AI model into the AI chip to determine a first performance value of the initial AI model based on the dataset;
(iv) determine a first probability that a current AI model should be replaced by the initial AI model, wherein the current AI model has a second performance value;
(v) determine, based on the first probability, whether to replace the current AI model with the initial AI model, by:
if the first probability has a value of one, determining that the current AI model be replaced by the initial AI model;
if the first probability has a value of less than one:
generating a random value;
comparing the random value to the first probability to determine whether to replace the current AI model with the initial AI model;

(vi) if it is determined that the current AI model be replaced with the initial AI model, replace the current AI model with the initial AI model;
(vii) transmit the current AI model and the first performance value of the initial AI model to the host device; and
repeat steps (iii-vii) for a number of iterations.

2. The device of claim 1 , wherein the initial AI model and the current AI model each includes a convolutional neural network (CNN) comprising a plurality of weights for loading into the AI chip.

3. The device of claim 1 , wherein programming instructions for determining the first probability comprise programming instructions configured to determine that the first probability has a value of one if the first performance value of the initial AI model has a higher value than the second performance value of the current AI model.

4. The device of claim 1 , wherein programming instructions for determining the first probability comprise programming instructions configured to determine the first probability based on a closeness of the first performance value of the initial AI model relative to the second performance value of the current AI model.

5. The device of claim 4 , wherein the first probability is also based on a current number of iterations.

6. The device of claim 1 further comprising additional programming instructions configured to update the initial AI model in each of the number of iterations.

7. A device comprising:
an artificial intelligence (AI) chip;
at least one processor; and
a non-transitory computer readable medium containing programming instructions that, when executed, will cause the at least one processor to:
(i) access a dataset;
(ii) receive an initial artificial intelligence (AI) model from a host device;
(iii) load the initial AI model into the AI chip to determine a first performance value of the initial AI model based on the dataset;
(iv) determine a first probability that a current AI model should be replaced by the initial AI model, wherein the current AI model has a second performance value;
(v) determine, based on the first probability, whether to replace the current AI model with the initial AI model;
(vi) if it is determined that the current AI model be replaced with the initial AI model, replace the current AI model with the initial AI model;
(vii) transmit the current AI model and the first performance value of the initial AI model to the host device;
repeat steps (iii-vii) for a number of iterations; and
update the initial AI model in each of the number of iterations,

wherein programming instructions for updating the initial AI model comprise programming instructions configured to:
determine a second probability of updating a first group of parameters of the initial AI model and an amplitude of change of parameters for the first group of parameters;
determine, based on the second probability, whether to update the first group of parameters of the initial AI model; and
if it is determined that the first group of parameters of the initial AI model be updated:
update the first group of parameters of the initial AI model by changing the first group of parameters by the amplitude of change;
otherwise, proceed to a next iteration without updating the first group of parameters of the initial AI model.

8. The device of claim 7 , wherein:
the initial AI model includes a convolutional neural network (CNN); and
the first group of parameters of the initial AI model include at least one of scalars, masks and biases of the CNN of the initial AI model.

9. A method comprising, by at least one processor:
(i) accessing a dataset;
(ii) receiving an initial artificial intelligence (AI) model from a host device;
(iii) loading the initial AI model into an AI chip to determine a first performance value of the initial AI model based on the dataset;
(iv) determining a first probability that a current AI model should be replaced by the initial AI model;
(v) determining, based on the first probability, whether to replace the current AI model with the initial AI model, by:
if the first probability has a value of one, determining that the current AI model be replaced by the initial AI model;
if the first probability has a value of less than one:
generating a random value;
comparing the random value to the first probability to determine whether to replace the current AI model with the initial AI model;

(vi) if it is determined that the current AI model be replaced with the initial AI model, replacing the current AI model with the initial AI model; and
(vii) transmitting the initial AI model and the first performance value of the initial AI model to the host device; and
repeating steps (iii-vii) for a number of iterations.

10. The method of claim 9 , wherein the initial AI model and the current AI model each includes a convolutional neural network (CNN) comprising a plurality of weights for loading into the AI chip.

11. The method of claim 9 , wherein determining the first probability comprises determining that the first probability has a value of one if the first performance value of the initial AI model has a higher value than the second performance value of the current AI model.

12. The method of claim 9 , wherein determining the first probability comprises determining the first probability based on a closeness of the first performance value of the initial AI model relative to the second performance value of the current AI model.

13. The device of claim 12 , wherein the first probability is also based on a current number of iterations.

14. The method of claim 9 further comprising updating the initial AI model in each of the number of iterations.

15. A method comprising, by at least one processor:
(i) accessing a dataset;
(ii) receiving an initial artificial intelligence (AI) model from a host device;
(iii) loading the initial AI model into an AI chip to determine a first performance value of the initial AI model based on the dataset;
(iv) determining a first probability that a current AI model should be replaced by the initial AI model;
(v) determining, based on the first probability, whether to replace the current AI model with the initial AI model;
(vi) if it is determined that the current AI model be replaced with the initial AI model, replacing the current AI model with the initial AI model;
(vii) transmitting the initial AI model and the first performance value of the initial AI model to the host device;
repeating steps (iii-vii) for a number of

Assignees

• Gyrfalcon Tech Inc

Inventors

• Zhang Yequn
• Ren Yongxiong
• Sun Baohua
• Yang Lin
• Dong Qi

Classifications

• G06N3/065

  Analogue means · CPC title

• G06N3/045

  Combinations of networks · CPC title

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06N3/098

  Distributed learning, e.g. federated learning · CPC title

• G06N3/09

  Supervised learning · CPC title