- ๋ฐํ์ค, 2023.01 ~
- (Spatial-Temporal) Graph Neural Networks๋ฅผ ๊ณต๋ถํฉ๋๋ค.
- Deep Learning์ ์ด์ฉํด ์ฐ๊ด๋ CV, NLP ๊น์ง์ ์์ฉ์ ๊ณต๋ถํฉ๋๋ค.
-
์ ๋ ํ์ฌ ์์ธ๋ํ๊ต ๊ณต๊ฐํต๊ณ์ฐ๊ตฌ์ค[Link]์์ ๊ณต๋ถ์ค์ ๋๋ค.
-
๊ณต๊ฐํต๊ณ(Spatial Statistics)๋
- As a sub-field of Statistics,
- Work on methodology for analyzing various spatial data or spatio-temporal data.
- Develop supporting theory of such methodology.
- Getting popular as more spatial and/or spatio-temporal data available.
- Data are spatially referenced so that each data point is associated with the location information.
-
์์ฌ ํ์ ๋ ผ๋ฌธ์ผ๋ก ์๊ณต๊ฐ๋ฐ์ดํฐ(Spatio-Temporal Data)์ ๋ํด ๋ค๋ฃจ๋ฉฐ Spatio-Temporal Data์ ํํ์ ๋ํด ์๊ฐํด๋ณด์์ต๋๋ค.
- To model spatial or spatio-temporal data, we assume that a set of data is a realization of a stochastic process.
- The most important characteristics of spatial or spatio-temporal data is dependence among data points incorporated with spatial or spatio-temporal information.
- ์๊ณต๊ฐ ๋ฐ์ดํฐ์ ์๊ณต๊ฐ์ ์์กด์ฑ์ Graphical Model๋ก ๋ฐ๋ผ๋ณด๋ คํฉ๋๋ค.
- ์ด๋ฅผ ์ํด Probabilistic Graphical Models์ ๋ํ ๊ณต๋ถ๊ฐ ํ์ํฉ๋๋ค.
- Bayesian Networks
- Probabilistic Graphical Models
- Pattern Recognition & Machien Learning, Bishop
- ์ด๋ฅผ ์ํด Recurrent Neural Networks (RNNs)์ ๋ํ ๊ณต๋ถ๊ฐ ํ์ํฉ๋๋ค.
- ์ด๋ฅผ ์ํด Graph Neural Networks (GNNs)์ ๋ํ ๊ณต๋ถ๊ฐ ํ์ํฉ๋๋ค.
- Overview