KR102086640B1 - User-definded machine learning apparatus for smart phone user and method thereof - Google Patents

Abstract

The present invention relates to a theft monitoring device and method for monitoring theft using a morphological feature, the user-defined machine learning apparatus, input unit for receiving data or commands from a user, user-defined machine learning using the received data A memory for storing a user-defined machine learning program that generates a model by learning a model or generates a result according to a user-defined machine learning model according to the received command, and at least one processor for driving the user-defined machine learning program. The user-defined machine learning program stored in the memory selects a predetermined machine learning algorithm for performing machine learning on the received data, and uses the selected machine learning algorithm to generate a user-defined machine learning model. Produce and the same species It is characterized by learning the generated user-defined machine learning model incrementally using the data of the smart phone users, by applying a variety of data to the machine learning model for various purposes, smart phone users can easily learn, and other data It can be applied for analysis.

Description

User-defined machine learning apparatus for smart phone user and method thereof

The present invention relates to a user-defined machine learning apparatus and method for allowing a smartphone user to freely input data to learn a machine learning model and to use the model, and a recording medium recording the method.

As the type and amount of data have soared, there have been many researches based on machine learning to find patterns from data and extract useful information from them. In particular, with the rapid increase in the number of smartphone users and the type and amount of data that can be obtained from the smartphone, applications that provide various services aimed at smartphone users are increasing exponentially. These applications focus on providing specific services that are targeted, and either use specific rules internally or use machine learning techniques. However, machine learning techniques are inherently applicable to various data and applications, so only the services provided by the application are different, and there are many cases in which the techniques used internally are not significantly different.

Korean Patent Laid-Open Publication No. 10-2007-0119105, December 20, 2007 Disclosure, People and People

The technical problem to be solved by the present invention is to provide a function to learn the machine learning model by receiving the data obtained on the smartphone, and to utilize in various data analysis.

In order to solve the above technical problem, a user-defined machine learning apparatus according to an embodiment of the present invention, the input unit for receiving data or commands from the user; A memory configured to learn and generate a user-defined machine learning model using the received data, or to store a user-defined machine learning program for deriving a result according to the user-defined machine learning model according to the received command; And at least one processor for driving the user-defined machine learning program, wherein the user-defined machine learning program stored in the memory selects a predetermined machine learning algorithm to perform machine learning on the input data, and The user-defined machine learning model is generated using the data through a selected machine learning algorithm, and the generated user-defined machine learning model is incrementally trained using the same kind of data.

According to an embodiment of the present invention, the data input by the input unit may be a user-defined machine learning device, characterized in that the data stored in the smart terminal or the data sensed by the sensor of the smart terminal.

According to an embodiment of the present invention, the data input by the input unit may be a user-defined machine learning device, characterized in that the data stored in the smart terminal or the data sensed by the sensor of the smart terminal.

According to an embodiment of the present invention, the user-defined machine learning program stored in the memory reads the user-defined machine learning model previously stored according to the received command, and uses the received data to read the user-defined machine. It may be a user-defined machine learning device characterized by incrementally learning the learning model.

According to an embodiment of the present invention, the machine learning algorithm selected by the user-defined machine learning program stored in the memory, the artificial neural network algorithm is the default (default), characterized in that can vary according to the user's selection It may be a learning device.

According to an embodiment of the present invention, the user-defined machine learning program stored in the memory may be a user-defined machine learning apparatus, characterized in that the machine learning algorithm selected according to the type of the received data is different.

According to an embodiment of the present invention, the user-defined machine learning program stored in the memory,

If the data received from the user is data for checking a result according to a previously stored user-defined machine learning model, the output for deriving a result according to a user-defined machine learning model for the data, and outputs the derived result It may be a user-defined machine learning device further comprising a wealth.

According to an embodiment of the present invention, the user machine further comprises a communication unit for transmitting or receiving a user-defined machine learning model, the user-defined machine learning model transmitted is characterized in that the user can learn in the other user terminal Definition May be a machine learning device.

In order to solve the above technical problem, the user-defined machine learning method according to an embodiment of the present invention, the user-defined machine learning system step of receiving data or command from the user; Selecting, by the user-defined machine learning system, a predetermined machine learning algorithm to perform machine learning on the received data; And generating, by the user-defined machine learning system, a user-defined machine learning model by performing machine learning using the data through the selected machine learning algorithm, wherein the user-defined machine learning model comprises the same kind of data. It is characterized by learning incrementally (increment) using.

According to an embodiment of the present invention, the data received from the user may be a user-defined machine learning method, characterized in that the data stored in the smart terminal or the data sensed by the sensor of the smart terminal.

According to an embodiment of the present invention, the user-defined machine learning system further comprises the step of reading a user-defined machine learning model stored in accordance with a command received from the user, the read using the received data It may be a user-defined machine learning method characterized by incrementally learning a user-defined machine learning model.

According to an embodiment of the present invention, the step of selecting the machine learning algorithm may be a user-defined machine learning method, characterized in that the machine learning algorithm selected according to the type of the received data is different.

According to an embodiment of the present invention, when the data input from the user is data for confirming a result according to a previously stored user-defined machine learning model, the user-defined machine learning model of the user-defined machine learning system for the data It may be a user-defined machine learning method further comprising providing a result according to the user.

According to an embodiment of the present invention, the user-defined machine learning system further comprises the step of transmitting the user-defined machine learning model to another user terminal, the transmitted user-defined machine learning model is learning in the other user terminal It may be a user-defined machine learning method characterized in that possible.

On the other hand, there is provided a computer-readable recording medium recording a program for executing the user-defined machine learning method on a computer.

According to embodiments of the present invention, by allowing smartphone users to easily apply a variety of data to a machine learning model for a variety of purposes for learning, and to apply for other data analysis, analysis of interest-oriented data from simple research or It can be used for various purposes, from analyzing data for investigation. In addition, by sharing the machine learning model between users can be expected to propagate and improve the useful machine learning model.

1 is a block diagram illustrating a user-defined machine learning apparatus according to an embodiment of the present invention.
2 is a diagram illustrating in detail a user-defined machine learning process according to an embodiment of the present invention.
3 is a flowchart illustrating a user-defined machine learning method according to an embodiment of the present invention.
4 to 6 are flowcharts illustrating a user-defined machine learning method according to an embodiment of the present invention.

Prior to the description of the specific contents of the present invention, for the convenience of understanding, the outline of the solution of the problem to be solved by the present invention or the core of the technical idea will be presented first.

In order for a smartphone user to freely generate and utilize a machine learning model, it is necessary to overcome the difficulties or limitations of applying data, machine learning algorithms, and machine learning methods. You can create and utilize your own machine learning models by providing a custom machine learning method that allows users to enter data and machine learning directly without setting up complex machine learning or prior knowledge of machine learning. have. Users can create machine learning models for a variety of purposes, allowing them to experience the learning and results of machine learning models using smartphone data, from simple interests to simple experiments. Further, by allowing each user to share and improve their own machine learning models, useful machine learning models can be propagated and improved by collective intelligence.

In order to provide such a user-defined machine learning apparatus, a user-defined machine learning apparatus according to an embodiment of the present invention, an input unit for receiving data or a command from a user, learning a user-defined machine learning model using the received data And a memory for storing a user-defined machine learning program for generating a result according to the user-defined machine learning model according to the command generated by the user, and at least one processor for driving the user-defined machine learning program. The user-defined machine learning program stored in the program selects a predetermined machine learning algorithm to perform machine learning on the received data, and generates a user-defined machine learning model using the data through the selected machine learning algorithm. The same kind of data And learning the generated user-defined machine learning model incrementally.

DETAILED DESCRIPTION Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. However, these examples are intended to illustrate the present invention in more detail, it will be apparent to those skilled in the art that the scope of the present invention is not limited thereby.

The configuration of the invention for clarifying the solution to the problem to be solved by the present invention will be described in detail with reference to the accompanying drawings, based on the preferred embodiment of the present invention, the same in the reference numerals to the components of the drawings The same reference numerals have been given to the components even though they are on different drawings, and it will be apparent that components of other drawings may be cited when necessary in describing the drawings. In addition, in describing the operation principle of the preferred embodiment of the present invention in detail, when it is determined that the detailed description of the known function or configuration and other matters related to the present invention may unnecessarily obscure the subject matter of the present invention, The detailed description is omitted.

1 is a block diagram illustrating a user-defined machine learning apparatus according to an embodiment of the present invention.

The user-defined machine learning apparatus 100 according to the embodiment of the present invention includes an input unit 10, a processor 20, and a memory 30, and may further include an output unit 40 or a communication unit 50. Can be.

The input unit 10 receives data or a command from a user.

The data received by the input unit 10 may be data stored in a smart terminal or data sensed by a sensor of the smart terminal. All data that can be generated in the smart terminal, such as time, GPS, WiFi, acceleration sensor, illuminance sensor, camera image may be included. Only the same data can be input as data for creating or training a user-defined machine learning model. By inputting only the same kind of data into the user-defined machine learning model, the user-defined machine learning model can be trained incrementally. The input data may be one kind or two or more kinds.

The data input from the user by the input unit 10 may further include a label for the data. Machine learning is divided into supervised, unsupervised, semi-supervised, and reinforcement (supervised, unsupervised, semi-supervised, reinforcement) .If machine learning, according to the user's purpose, is supervised learning, it is labeled with data. Can be entered. When the labels are input together, it may be determined as data for supervised learning, and when the labels are not input together, it may be determined as unsupervised learning.

The memory 30 stores a user-defined machine learning program that learns and generates a user-defined machine learning model using the input data or derives a result according to the user-defined machine learning model according to the input command. In addition, the memory 30 stores the generated machine learning model.

The processor 20 drives the user-defined machine learning program.

The user-defined machine learning program stored in the memory selects a predetermined machine learning algorithm to perform machine learning on the input data, and generates a user-defined machine learning model using the data through the selected machine learning algorithm. And learning the generated user-defined machine learning model incrementally using the same kind of data.

When there is no initial user-defined machine learning model or when a new user-defined machine learning model is created, a machine learning algorithm is selected to perform machine learning on the input data. The machine learning algorithm preferably selects an algorithm that can be incrementally learned. When the user inputs data for learning the machine learning model, the whole learning of the whole data becomes less efficient. Therefore, the machine learning algorithm uses an algorithm that can be incrementally learned whenever the data is newly input.

In the machine learning algorithm selected by the user-defined machine learning program stored in the memory, the artificial neural network algorithm is the default and may vary according to a user's selection. In addition to artificial neural network algorithms, all other machine learning algorithms such as classification, regression, and clustering can be used.

The machine learning algorithm selected by the user-defined machine learning program stored in the memory may be selected according to the type of the received data. Depending on the type of data, the appropriate machine learning algorithm can be selected. For example, when the type of data to be input is an image, a convolutional neural network (CNN) may be selected. The machine learning algorithm selected according to the type of data may be preset. This may be set by a user-defined machine learning program provider or by using information that has been selected for the same kind of data in the past. Of course, even if it is set in advance, it can be changed by the user.

If there is a user-defined machine learning model previously stored in the memory 30, and the user inputs a command for learning the user-defined machine learning model stored by inputting data, the user-defined machine learning stored in the memory. The program reads a user-defined machine learning model previously stored according to the received command, and incrementally learns the read user-defined machine learning model using the received data. At this time, the received data should be the same kind of data as the data input as the input data of the read user-defined machine learning model. By incrementally learning the read user-defined machine learning model using the input data, it is possible to learn more precise and accurate machine learning model.

If there is a user-defined machine learning model previously stored in the memory 30, and the user inputs a command for confirming a result according to the user-defined machine learning model stored in the user by inputting data, the user-defined user learning stored in the memory The machine learning program derives a result according to a user-defined machine learning model for the received data.

In this case, the derived result may be provided to the user by being output to the output unit 40. The output unit 40 may be a display screen.

Not only does the user create and utilize a user-defined machine learning model using data of his smart terminal, but also by spreading useful machine learning models by sharing user-defined machine learning models with other users, Improvements in learning models can also be expected. To this end, the communication unit 50 may transmit the user-defined machine learning model to another user terminal. The transmitted user-defined machine learning model may be learned in the other user terminal, so that the user-defined machine learning model may be improved by using data input from the other user terminal. If only the propagation of the user-defined machine learning model is desired, it may be encrypted to prevent further learning or modification at other user terminals.

FIG. 2 is a diagram illustrating a user-defined machine learning process in detail according to an embodiment of the present invention, and will be described with reference to an embodiment embodied as an application that can be utilized by a user.

If an application is provided that allows users to easily analyze the data they get from their smartphones, they can use it for a variety of purposes. Using the data available from the smartphone, you can learn any machine learning model on your smartphone and use the learned model to analyze new data. Data analysis for a variety of purposes, from simple data analysis to research. For example, as an interesting data analysis, enter your friends 'face photos as' ugly ', learn the actors' face photos as 'handsome', and then put other people's face photos to see the results. You can also observe mainly. Furthermore, users can transfer or share the different types of machine learning models created by users, and further improve data by adding data to the models.

The main screen displays the name of the current machine learning model. There are five choices and you can choose one of them. The name of the machine learning model here is not the name of the machine learning algorithm itself, but an arbitrary name given by the user. If you do not load an existing model with item 2 and do not specify a different name, the machine learning model is set to an arbitrary default name (eg noname_01).

When the user selects item 1, data can be inputted into the current machine learning model, and the model performs incremental learning using the inputted data. The data can be selected from the photos, videos, text, time, etc., and the data available from various sensors (illuminance sensor, acceleration sensor, GPS, WiFi sensor, etc.). However, each model may receive only the same type of data afterwards as the type of data initially input. For example, if you entered an image as the first training data for a new model, you can only enter image data in this model. Of course, a plurality of data types may be received when data is input. For example, you can input image and light sensor data as the first training data of a new model, and then you need to input image and light sensor data in the same way.

The machine learning model can be used in various ways such as artificial neural network model, SVM, etc., but it should be a model that can be incremental learning, basically it is set as artificial neural network model, and user can choose to use another model separately Do.

If the user selects item 2, the model that was previously saved by name 3 can be loaded or deleted. Alternatively, you can delete the current model and create a new one.

If the user selects item 3, the user can give the model an arbitrary name and save it.

If the user selects item 4, input new data into the current machine learning model, but not for training, but for analysis. In other words, the data input through item 4 is not used to train the model, so that the model can be provided with the analysis results. Of course, the same type of data that was used to train the model must be entered.

If the user selects item 5, various settings of the application can be made. If several machine learning models are provided, the type of model can be selected and various parameters of the machine learning model can be modified. We will not use item 5 for interesting data analysis, but if you use it for simple research or research purposes, you will often use it.

In addition, it provides the ability to easily send and receive machine learning models between users. For example, if user A sends machine learning model K to user B, user B may receive this model and then load this model through item 2 on the user menu. By inserting additional data into the loaded model, you can create an improved model K 'and propagate it back to other users.

Assuming that a user generates a machine learning model for an image through the process of generating a user-defined machine learning model, the following process may be described.

(1) The user took five male face pictures and five female face pictures, and let them be Dm and Dw, respectively.

(2) Run the application of the present invention, select 'Model Training' to create a new model and name it K.

(3) When entering data into the model, enter Dm 5 with the label 'Man' and Dw 5 with the label 'Woman'.

(4) Model training is performed, and feature vector Fi is extracted from each image i, and arbitrary model training is performed based on this feature vector and label.

(5) It is not limited to specific feature and model at the time of learning. However, assuming that the Convolutional Neural Network (CNN) is known to be effective in various images and texts, the feature will be the pixel RGB value of the image itself. (The user only labels each image Man and Woman, but nothing else.)

(6) Save the trained model through the 'model storage' function.

(7) Enter a new picture through the 'use model' function. After entering 'man's face', you can see that 'Man' is the result.

(8) Send the model to other friends by SNS or e-mail, or send the model (model parameter) to a friend using the 'Send Model' function in the application. When using a CNN, the values of the weight matrix parameters of the CNN will be transmitted.

(9) The friend who received the model reads the model with the 'model loading' function, and then adds a new male face picture and a female face picture to improve the model.

(10) The improved version can be shared again with other friends.

3 is a flowchart illustrating a user-defined machine learning method according to an embodiment of the present invention, and FIGS. 4 to 6 are flowcharts illustrating a user-defined machine learning method according to an embodiment of the present invention. 3 to 6 correspond to the description of the user-defined machine learning apparatus and its process of FIGS. 1 to 2, and thus redundant descriptions thereof will be omitted.

In step S110 the user-defined machine learning system receives data or commands from the user. The data received from the user may be data stored in a smart terminal or data sensed by a sensor of the smart terminal.

In step S120 the user-defined machine learning system selects a predetermined machine learning algorithm to perform machine learning on the received data. The machine learning algorithm selected according to the type of data input in step S110 may vary.

In operation S130, the user-defined machine learning system generates a user-defined machine learning model by performing machine learning using the data through the selected machine learning algorithm. In this case, the user-defined machine learning model is learned incrementally using the same kind of data.

In the case of learning the previously stored user-defined machine learning model, in step S210, the user-defined machine learning system reads the user-defined machine learning model stored in accordance with the command input from the user, and in step S110 Can be used to incrementally train the read-out user-defined machine learning model.

If the data received from the user is data for confirming the result according to the previously stored user-defined machine learning model, in step S310, the user-defined machine learning system to the result according to the user-defined machine learning model for the data Derived and provided to the user.

In order to share and improve the machine learning model between users, the user-defined machine learning system transmits the user-defined machine learning model to another user terminal in step S410, and the user-defined machine learning model transmitted is learned in the other user terminal. By making this possible, it is possible to improve the model.

Method according to an embodiment of the present invention can be implemented in the form of program instructions that can be executed by various computer means may be recorded on a computer readable medium. The computer readable medium may include program instructions, data files, data structures, and the like, alone or in combination. Program instructions recorded on the media may be those specially designed and constructed for the purposes of the present invention, or they may be of the kind well-known and available to those having skill in the computer software arts. Examples of computer readable recording media include magnetic media such as hard disks, floppy disks and magnetic tape, optical media such as CD-ROMs, DVDs, and magnetic disks such as floppy disks. Magneto-optical media, and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, flash memory, and the like. Examples of program instructions include not only machine code generated by a compiler, but also high-level language code that can be executed by a computer using an interpreter or the like. The hardware device may be configured to operate as one or more software modules to perform the operations of the present invention, and vice versa.

In the present invention as described above has been described by the specific embodiments, such as specific components and limited embodiments and drawings, but this is provided to help a more general understanding of the present invention, the present invention is not limited to the above embodiments. For those skilled in the art, various modifications and variations are possible from these descriptions.

Accordingly, the spirit of the present invention should not be limited to the described embodiments, and all of the equivalents and equivalents of the claims, as well as the appended claims, will fall within the scope of the present invention. .

100: custom machine learning device
10: input unit
20: processor
30: memory
40: output unit
50: communication unit

Claims (15)

An input unit configured to receive data or a command of the same type from a user;
A memory configured to learn and generate a user-defined machine learning model using the input data or to store a user-defined machine learning program for deriving a result according to the user-defined machine learning model according to the received command;
At least one processor for driving the user-defined machine learning program; And
Including a communication unit for transmitting a user-defined machine learning model to another user terminal,
The user-defined machine learning program stored in the memory,
Select a predetermined machine learning algorithm to perform the machine learning on the input data, create a user-defined machine learning model using the data through the selected machine learning algorithm, the data using the same type of data Each time a new input is received, the generated user-defined machine learning model is learned incrementally, and the user-defined machine learning model is transmitted and shared with other users to share the same type of data newly input by the other user terminal. A user-defined machine learning apparatus, characterized in that to improve the user-defined machine learning model. The method of claim 1,
The data input by the input unit,
User-defined machine learning device, characterized in that the data stored in the smart terminal or data sensed by the sensor of the smart terminal. The method of claim 1,
The data received by the input unit is,
And a label for the data. The method of claim 1,
The user-defined machine learning program stored in the memory,
The user-defined machine learning apparatus according to the received command, characterized in that for reading the previously stored user-defined machine learning model, and incrementally learning the read user-defined machine learning model using the received data. The method of claim 1,
Machine learning algorithm selected by the user-defined machine learning program stored in the memory,
The artificial neural network algorithm is the default and can be changed according to a user's selection. The method of claim 1,
The user-defined machine learning program stored in the memory,
And a machine learning algorithm selected according to the type of the received data. The method of claim 1,
The user-defined machine learning program stored in the memory,
If the data received from the user is data for confirming the result according to the previously stored user-defined machine learning model, to derive the result according to the user-defined machine learning model for the data,
User-defined machine learning apparatus further comprises an output unit for outputting the derived result. delete Receiving, by the user-defined machine learning system, the same kind of data or command from the user;
Selecting, by the user-defined machine learning system, a predetermined machine learning algorithm to perform machine learning on the received data;
Generating, by the user-defined machine learning system, a user-defined machine learning model by performing machine learning using the data through the selected machine learning algorithm; And
And transmitting, by the user-defined machine learning system, the user-defined machine learning model to another user terminal.
The user-defined machine learning model,
A user characterized by learning incrementally each time data is newly input using the same type of data, and being shared with other users, thereby improving by using the same type of data newly input from the other user terminal. Definition of machine learning methods. The method of claim 9,
Data received from the user,
User-defined machine learning method characterized in that the data stored in the smart terminal or data sensed by the sensor of the smart terminal. The method of claim 9,
The user-defined machine learning system further comprises the step of reading a user-defined machine learning model stored in accordance with the command received from the user,
And a method for incrementally learning the read user-defined machine learning model using the input data. The method of claim 9,
Selecting the machine learning algorithm,
User-defined machine learning method, characterized in that the machine learning algorithm is selected according to the type of the received data. The method of claim 9,
When the data input from the user is data for confirming a result according to a previously stored user-defined machine learning model, the user-defined machine learning system may provide a user with a result according to a user-defined machine learning model for the data. Custom machine learning method further comprising a step. delete A computer-readable recording medium having recorded thereon a program for executing the method of any one of claims 9 to 13.

Images