US 11,868,904 B2
Prediction model training management system, method of the same, master apparatus and slave apparatus for the same
Choong Seon Hong, Yongin-si (KR); Thar Kyi, Yongin-si (KR); and Do Hyun Kim, Suwon-si (KR)
Assigned to University-Industry Cooperation Group of Kyung-Hee University, Yongin-si (KR)
Filed by University-Industry Cooperation Group of Kyung-Hee University, Yongin-si (KR)
Filed on Oct. 26, 2018, as Appl. No. 16/172,399.
Claims priority of application No. 10-2018-0122683 (KR), filed on Oct. 15, 2018.
Prior Publication US 2020/0118007 A1, Apr. 16, 2020
Int. Cl. G06N 5/02 (2023.01); G06N 3/04 (2023.01); G06N 3/08 (2023.01)
CPC G06N 5/02 (2013.01) [G06N 3/04 (2013.01); G06N 3/08 (2013.01)] 14 Claims
OG exemplary drawing
 
1. A system for training and managing a prediction model, the system comprising:
a master apparatus comprising a hardware processor and a hardware storage,
wherein the hardware processor of the master apparatus is configured to
generate a prediction model,
train the prediction model, and
obtain the trained prediction model; and
a slave apparatus comprising a hardware processor and a hardware storage,
wherein the hardware processor of the slave apparatus is configured to collect data,
transmit the data to the master apparatus,
receive the prediction model or the trained prediction model from the master apparatus, and
operate based on the prediction model or the trained prediction model,
wherein the hardware processor of the master apparatus is further configured to generate the prediction model or train the prediction model based on the data transmitted from the slave apparatus,
wherein the prediction model comprises:
a first prediction model configured to obtain a prediction result about a class corresponding to input data; and
a second prediction model configured to receive a result of the first prediction model and obtain a prediction result corresponding to the result of the first prediction model,
wherein the first prediction model comprises a first algorithm to which first data is input and a second algorithm to which an output result of the first algorithm is input,
wherein the hardware storage of the master apparatus is configured to store feedback information that includes information on training accuracy and learning loss of a respective model, and
wherein:
the first algorithm is implemented using a convolutional neural network (CNN), and the second algorithm is implemented using a recurrent neural network (RNN), which is designed for temporal dynamics of input sequential data and includes a plurality of RNN cells;
the first data is inputted to and then convolution-processed by the CNN of the first algorithm, second data, which is convolution result data, is acquired, and the second data is input to the RNN of the second algorithm; and
third data, which is output data of the RNN, is transferred to the second prediction model.