ARTIFICIAL INTELLIGENCE SERVER

An artificial intelligence (AI) server is provided. The AI server includes a communication interface configured to communicate with an electronic device, and at least one processor configured to update a classification layer by training an artificial intelligence model in such a manner that classification training data and classification labeling data are provided to the artificial intelligence model including a feature extraction layer for extracting a feature vector and a classification layer for classifying input data using the feature vector, and transmit the updated classification layer to the electronic device.

1 . An artificial intelligence server comprising:

a communication interface configured to communicate with an electronic device; and

at least one processor configured to:

update a classification layer by training an artificial intelligence model in such a manner that classification training data and classification labeling data are provided to the artificial intelligence model including a feature extraction layer for extracting a feature vector and a classification layer for classifying input data using the feature vector; and

transmit the updated classification layer to the electronic device.

2 . The artificial intelligence server according to
claim 1
, wherein the at least one processor is configured to generate the artificial intelligence model including the feature extraction layer having a first parameter and the classification layer having a (2-1) th parameter by training a neural network using general-purpose training data and general-purpose labeling data.

3 . The artificial intelligence server according to
claim 2
, wherein the at least one processor is configured to obtain the classification layer having a (2-2) th parameter different from the (2-1) th parameter by training the artificial intelligence model in such a manner that the classification training data and the classification labeling data are provided to the artificial intelligence model.

4 . The artificial intelligence server according to
claim 3
, wherein the at least one processor is configured to:

replace the classification layer having the (2-1) th parameter with a classification layer having an initial parameter; and

obtain the classification layer having the (2-2) th parameter by training the artificial intelligence model in such a manner that the classification training data and the classification labeling data are provided to the artificial intelligence model.

5 . The artificial intelligence server according to
claim 3
, wherein the at least one processor is configured to:

adjust a parameter of the feature extraction layer and a parameter of the classification layer, so that an error between an estimated value of the neural network and the general-purpose labeling data is reduced, in the process of training the neural network; and

adjust the parameter of the classification layer, so that an error between an estimated value of the artificial intelligence model and the classification labeling data is reduced, in the process of training the artificial intelligence model.

6 . The artificial intelligence server according to
claim 3
, wherein, even after the artificial intelligence model is trained, the first parameter of the feature extraction layer remains as before the training of the artificial intelligence model.

7 . The artificial intelligence server according to
claim 3
, wherein the trained artificial intelligence model comprises the feature extraction layer having the first parameter and the classification layer having the (2-2) th parameter, and

wherein the at least one processor is configured to:

separate the classification layer having the (2-2) th parameter from the feature extraction layer; and

transmit the separated classification layer to the electronic device.

8 . The artificial intelligence server according to
claim 1
, further comprising at least one memory configured to store a plurality of classification layers having different parameters,

wherein the at least one processor is configured to transmit, to the electronic device, one or more classification layers among the plurality of classification layers if a classification layer change request is received from the electronic device.

9 . The artificial intelligence server according to
claim 1
, further comprising at least one memory configured to store data,

wherein the at least one processor is configured to:

obtain a classification layer corresponding to a first category by updating the classification layer by providing classification training data and classification labeling data corresponding to the first category to the artificial intelligence model, and store the classification layer corresponding to the first category in the at least one memory; and

obtain a classification layer corresponding to a second category by updating the classification layer by providing classification training data and classification labeling data corresponding to the second category to the artificial intelligence model, and store the classification layer corresponding to the second category in the at least one memory.

10 . The artificial intelligence server according to
claim 9
, wherein the at least one processor is configured to:

receive category selection information from the electronic device; and

transmit, to the electronic device, the classification layer corresponding to the first category if the received category selection information corresponds to the first category.

11 . An operating method of an artificial intelligence server, the operating method comprising:

updating a classification layer by training an artificial intelligence model in such a manner that classification training data and classification labeling data are provided to the artificial intelligence model including a feature extraction layer for extracting a feature vector and a classification layer for classifying input data using the feature vector; and

transmitting the updated classification layer to an electronic device.

12 . The operating method according to
claim 11
, further comprising generating the artificial intelligence model including the feature extraction layer having a first parameter and the classification layer having a (2-1) th parameter by training a neural network using general-purpose training data and general-purpose labeling data.

13 . The operating method according to
claim 12
, wherein the updating of the classification layer comprises obtaining the classification layer having a (2-2) th parameter different from the (2-1) th parameter by training the artificial intelligence model in such a manner that the classification training data and the classification labeling data are provided to the artificial intelligence model.

14 . The operating method according to
claim 13
, wherein the obtaining of the classification layer having the (2-2) th parameter comprises:

replacing the classification layer having the (2-1) th parameter with a classification layer having an initial parameter; and

obtaining the classification layer having the (2-2) th parameter by training the artificial intelligence model in such a manner that the classification training data and the classification labeling data are provided to the artificial intelligence model.

15 . The operating method according to
claim 13
, wherein the generating of the artificial intelligence model comprises adjusting a parameter of the feature extraction layer and a parameter of the classification layer, so that an error between an estimated value of the neural network and the general-purpose labeling data is reduced, in the process of training the neural network, and

wherein the updating of the classification layer comprises adjusting the parameter of the classification layer, so that an error between an estimated value of the artificial intelligence model and the classification labeling data is reduced, in the process of training the artificial intelligence model.

16 . The operating method according to
claim 13
, wherein, even after the artificial intelligence model is trained, the first parameter of the feature extraction layer remains as before the training of the artificial intelligence model.

17 . The operating method according to
claim 13
, wherein the trained artificial intelligence model comprises the feature extraction layer having the first parameter and the classification layer having the (2-2) th parameter, and

wherein the transmitting of the updated classification layer to the electronic device comprises:

separating the classification layer having the (2-2) th parameter from the feature extraction layer; and

transmitting the separated classification layer to the electronic device.

18 . The operating method according to
claim 11
, wherein the transmitting of the updated classification layer to the electronic device comprises transmitting, to the electronic device, one or more classification layers among a plurality of classification layers having different parameters if a classification layer change request is received from the electronic device.

19 . The operating method according to
claim 11
, wherein the updating of the classification layer comprises:

obtaining a classification layer corresponding to a first category by updating the classification layer by providing classification training data and classification labeling data corresponding to the first category to the artificial intelligence model, and storing the classification layer corresponding to the first category in the at least one memory; and

obtaining a classification layer corresponding to a second category by updating the classification layer by providing classification training data and classification labeling data corresponding to the second category to the artificial intelligence model, and storing the classification layer corresponding to the second category in the at least one memory.

20 . The operating method according to
claim 19
, wherein the transmitting of the updated classification layer to the electronic device comprises:

receiving category selection information from the electronic device; and

transmitting, to the electronic device, the classification layer corresponding to the first category if the received category selection information corresponds to the first category.

CROSS-REFERENCE TO RELATED APPLICATIONS
[0001]
This application claims priority to Korean Patent Application No. 10-2019-0138821 filed in the Republic of Korea on Nov. 1, 2019, the entire contents of which are hereby incorporated by reference in its entirety.

BACKGROUND
[0002]
The present disclosure relates to an artificial intelligence server capable of updating and providing a classification layer among a feature extraction layer and the classification layer constituting an artificial intelligence model.

[0003]
Artificial intelligence (AI) is one field of computer engineering and information technology for studying a method of enabling a computer to perform thinking, learning, and self-development that can be performed by human intelligence and may denote that a computer imitates an intelligent action of a human.

[0004]
Also, AI is directly or indirectly associated with the other field of computer engineering without being individually provided. Particularly, at present, in various fields of information technology, an attempt to introduce AI components and use the AI components in solving a problem of a corresponding field is being actively done.

[0005]
Meanwhile, techniques for perceiving and learning the surrounding situation by using AI and providing information desired by the user in a desired form, or performing an operation or function desired by the user are being actively studied.

[0006]
An electronic device that provides such various operations and functions may be referred to as an AI device.

[0007]
A model learned through machine learning cannot depart from the learning range during inference.

[0008]
For example, in the case of face recognition, when training is performed using white-oriented training data, it is obvious that the inference accuracy of Asians is inevitably deteriorated due to the difference with the training data.

[0009]
If the entire AI model is retrained with Asian-oriented training data so as to ensure the inference accuracy for Asians, the same costs as when previously trained occur again.

[0010]
Also, a general-purpose model training for whites and Asians may not be able to perform performance that meets user requirements. For example, if a user only uses an AI model to classify whites, a white-optimized model may exhibit better performance than a general-purpose model.

SUMMARY
[0011]
The present disclosure has been made in an effort to solve the above-described problems and provides an AI server capable of updating and providing a classification layer among a feature extraction layer and the classification layer constituting an AI model.

[0012]
In one embodiment, an AI server includes a communication interface configured to communicate with an electronic device, and at least one processor configured to update a classification layer by training an artificial intelligence model in such a manner that classification training data and classification labeling data are provided to the artificial intelligence model including a feature extraction layer for extracting a feature vector and a classification layer for classifying input data using the feature vector, and transmit the updated classification layer to the electronic device.

[0013]
The details of one or more embodiments are set forth in the accompanying drawings and the description below. Other features will be apparent from the description and drawings, and from the claims.

BRIEF DESCRIPTION OF THE DRAWINGS
[0014]
FIG. 1 illustrates an AI device 100 according to an embodiment of the present disclosure.

[0015]
FIG. 2 illustrates an AI server 200 according to an embodiment of the present disclosure.

[0016]
FIG. 3 illustrates an AI system 1 according to an embodiment of the present disclosure.

[0017]
FIG. 4 is a view for describing an operating method of an AI server.

[0018]
FIG. 5 is a view for describing a convolutional neural network (CNN) among neural networks.

[0019]
FIG. 6 is a view for describing a method of generating an AI model by training a CNN.

[0020]
FIG. 7 is a view for describing distribution of the generated AI model.

[0021]
FIG. 8 is a view for describing a method of updating a classification layer.

[0022]
FIG. 9 is a view for describing distribution of the classification layer.

[0023]
FIG. 10 is a view for describing a method of generating a plurality of classification layers and providing the plurality of classification layers to an electronic device.

[0024]
FIG. 11 is a view for describing a method of replacing a classification layer.

[0025]
FIG. 12 is a view for describing a method of providing a classification layer of a category desired by a user.

[0026]
FIG. 13 is a view for describing another method of providing a classification layer of a category desired by a user.

DETAILED DESCRIPTION OF THE EMBODIMENTS
[0027]
Hereinafter, embodiments of the present disclosure are described in more detail with reference to accompanying drawings and regardless of the drawings symbols, same or similar components are assigned with the same reference numerals and thus overlapping descriptions for those are omitted. The suffixes “module” and “unit” for components used in the description below are assigned or mixed in consideration of easiness in writing the specification and do not have distinctive meanings or roles by themselves. In the following description, detailed descriptions of well-known functions or constructions will be omitted since they would obscure the disclosure in unnecessary detail. Additionally, the accompanying drawings are used to help easily understanding embodiments disclosed herein but the technical idea of the present disclosure is not limited thereto. It should be understood that all of variations, equivalents or substitutes contained in the concept and technical scope of the present disclosure are also included.

[0028]
It will be understood that the terms “first” and “second” are used herein to describe various components but these components should not be limited by these terms. These terms are used only to distinguish one component from other components.

[0029]
In this disclosure below, when one part (or element, device, etc.) is referred to as being ‘connected’ to another part (or element, device, etc.), it should be understood that the former can be ‘directly connected’ to the latter, or ‘electrically connected’ to the latter via an intervening part (or element, device, etc.). It will be further understood that when one component is referred to as being ‘directly connected’ or ‘directly linked’ to another component, it means that no intervening component is present.

[0030]

[0031]
Artificial intelligence refers to the field of studying artificial intelligence or methodology for making artificial intelligence, and machine learning refers to the field of defining various issues dealt with in the field of artificial intelligence and studying methodology for solving the various issues. Machine learning is defined as an algori…