Selecting artificial intelligence model based on input data

Patent metadata

Field	Value
Publication number	US-11676016-B2
Application number	US-202016899841-A
Country	US
Kind code	B2
Filing date	Jun 12, 2020
Priority date	Jun 12, 2019
Publication date	Jun 13, 2023
Grant date	Jun 13, 2023

Abstract

A method of selecting an artificial intelligence (AI) model based on input data to select an AI model capable of correctly obtaining a result corresponding to data to be classified, e.g., a classification result indicating one of at least one class, from among a plurality of AI models, and a display device for performing the method.

First claim

What is claimed is:

1. A method of selecting an artificial intelligence model for classifying input data, the method comprising:
obtaining first misclassified data of a first artificial intelligence model classifying data included in a training data set, the first misclassified data indicating data misclassified by the first artificial intelligence model from the training data set misclassified data corresponding to misclassification results of a training data set;
obtaining second misclassified data of a second artificial intelligence model classifying the data included in the training data set, the second misclassified data indicating data misclassified by the second artificial intelligence model from the training data set misclassified data corresponding to misclassification results of a training data set;
training a selection artificial intelligence model based on the first misclassified data and the second misclassified data;
outputting a first misclassification probability of the first artificial intelligence model incorrectly classifying input data and a second misclassification probability of the second artificial intelligence model incorrectly classifying the input data, in response to the input data being received in the trained selection artificial intelligence model; and
selecting an optimal artificial intelligence model for the input data from among the first artificial intelligence model and the second artificial intelligence model based on the first misclassification probability and the second misclassification probability.

2. The method of claim 1 , wherein the selecting comprises selecting the optimal artificial intelligence model corresponding to a lowest value of misclassification probability of incorrectly classifying the input data determined by the selection artificial intelligence model from among the first misclassification probability and the second misclassification probability.

3. The method of claim 1 , wherein the training comprises:
inputting the first misclassified data and the second misclassified data to the selection artificial intelligence model.

4. The method of claim 3 , wherein the training comprises training the selection artificial intelligence model about characteristics and distribution of data of the first data and the second data.

5. The method of claim 1 , further comprising:
providing the input data to the optimal artificial intelligence model; and
obtaining a result of classifying the input data from the optimal artificial intelligence model.

6. The method of claim 1 , wherein a neural network structure of the selection artificial intelligence model is different from structures of the first artificial intelligence model and the second artificial intelligence model.

7. The method of claim 1 , further comprising adjusting, based on the first misclassification probability and the second misclassification probability, a plurality of weight values applied to at least one of the first artificial intelligence model and the second artificial intelligence model comprising an ensemble artificial intelligence model.

8. The method of claim 7 , wherein the adjusting of the plurality of weight values comprises adjusting the plurality of weight values applied to at least one of the first artificial intelligence model and the second artificial intelligence model, in such a manner that the plurality of weight values are reduced when misclassification probability is high.

9. The method of claim 7 , further comprising:
providing the input data to the ensemble artificial intelligence model; and
outputting a result of classifying the input data from the ensemble artificial intelligence model.

10. The method of claim 1 , wherein the first artificial intelligence model is configured to classify the input data into a class from among a plurality of classes and the second artificial intelligence model is configured to classify the input data into the class from among the plurality of classes.

11. The method of claim 1 , wherein at least one of a first hyper-parameter, a first model architecture, a first training technique, or a first dataset input for training the first artificial intelligence model is different from at least one of a second hyper-parameter, a second model architecture, a second training technique, or a second dataset input for training the second artificial intelligence model.

12. The method of claim 1 , wherein the first artificial intelligence model and the second artificial intelligence model are combined as an ensemble AI model configured to classify the input data.

13. The method of claim 1 , wherein the first artificial intelligence model comprises at least one on-device artificial intelligence model and at least one server-based artificial intelligence model.

14. A device for selecting an artificial intelligence model for classifying input data, the device comprising:
a display; and
a processor configured to execute at least one instruction to:
obtain first misclassified data of a first artificial intelligence model classifying data included in a training data set, the first misclassified data indicating data misclassified by the first artificial intelligence model from the training data set misclassified data corresponding to misclassification results of a training data set;
obtain second misclassified data of a second artificial intelligence model classifying the data included in the training data set, the second misclassified data indicating data misclassified by the second artificial intelligence model from the training data set misclassified data corresponding to misclassification results of a training data set;
train a selection artificial intelligence model based on the first misclassified data and the second misclassified data;
output a first misclassification probability of the first artificial intelligence model incorrectly classifying input data and a second misclassification probability of the second artificial intelligence model incorrectly classifying the input data, in response to the input data being received in the trained selection artificial intelligence model; and
select an optimal artificial intelligence model for the input data from among the first artificial intelligence model and the second artificial intelligence model based on the first misclassification probability and the second misclassification probability.

15. The device of claim 14 , wherein the processor is further configured to select the optimal artificial intelligence model corresponding to a lowest value of misclassification probability of incorrectly classifying the input data determined by the selection artificial intelligence model from among the first misclassification probability and the second misclassification probability.

16. The device of claim 14 , wherein the processor is further configured to input the first misclassified data and the second misclassified data to the selection artificial intelligence model.

17. The device of claim 14 , wherein the first artificial intelligence model is configured to classify the input data into a class from among a plurality of classes and the second artificial intelligence model is configured to classify the input data into the class from among the plurality of classes.

18. The device of claim 14 , further comprising a communicator configured to communicate with a server,
wherein the processor is further configured to control the communicator to receive the first misclassified data and the second misclassified data from the server.

19. A non-tra

Assignees

• Samsung Electronics Co Ltd

Inventors

• Baijal Anant
• Agarwal Vivek
• Koo Jayoon

Classifications

• G06N3/08 Primary

  Learning methods · CPC title

• G06N20/00

  Machine learning · CPC title

• G06F3/14

  Digital output to display device {; Cooperation and interconnection of the display device with other functional units} · CPC title

• G06N7/01 Primary

  Probabilistic graphical models, e.g. probabilistic networks · CPC title

• G06N20/20

  Ensemble learning · CPC title