APPARATUS AND METHOD FOR CLASSIFICATION OF SINGLE MOLECULE PHASES USING GRAPH CONVOLUTIONAL NETWORK
An operating method of a molecular phase classification device comprises: a step of generating a feature vector based on the position information of molecules inputted as learning data, generating a neighbor matrix based on the connection information between the molecules, and converting the positio...
Saved in:
Main Authors | , |
---|---|
Format | Patent |
Language | English Korean |
Published |
16.12.2022
|
Subjects | |
Online Access | Get full text |
Cover
Loading…
Summary: | An operating method of a molecular phase classification device comprises: a step of generating a feature vector based on the position information of molecules inputted as learning data, generating a neighbor matrix based on the connection information between the molecules, and converting the position information into the graph data formed by the feature vector and the neighbor matrix; and a step of learning a graph convolution model to classify the molecular phase corresponding to the graph data. Therefore, the present invention is capable of more accurately classifying the molecular phases.
분자 상 분류 장치의 동작 방법으로서, 학습데이터로 입력된 분자들의 위치 정보에 기초하여 특징 벡터를 생성하고, 분자 사이의 연결 정보에 기초하여 이웃행렬을 생성하여, 특징 벡터와 이웃행렬로 형성된 그래프 데이터로 변환하는 단계, 그리고 그래프 데이터에 대응하는 분자 상을 분류하도록 그래프 컨볼루션 모델을 학습시키는 단계를 포함한다. |
---|---|
Bibliography: | Application Number: KR20210074929 |