13.3. Estimating MLE parameters by using Kalman filter

13.3. Estimating MLE parameters by using Kalman filter
전통적인 MLE방식의 parameter 추론은 전체 data를 입력한 likelihood를 극대화하는 parameter를 one time에 찾아낸다. 본 영상에서는, Kalman filter를 활용한다면 data가 누적될 때 마다 새로운 likelihood함수를 구현하고, 해당 신규 data를 활용하여 매번 parameter를 업데이트하는 방식을 소개한다.
Conventional MLE approach inputs all the available data into likelihood function which is maximized by optimal parameters. Using Kalman filter, this video shows that whenever new data comes in, we can establish new likelihood function containing the components of state-space representation of a given time-series. Then we find optimal parameters maximizing the likelihood. Once we encounter new data, we re-optimize the parameters maximizing new likelihood. In conclusion, Kalman filter helps us with sequentially update optimal parameters of MLE, while conventional approach allows one-time optimization.
(Important) The numerical example presented during the lecture is original work of the instructor and does not come from the reference book below.
Click the following link to get the lecture note presented during the video.
drive.google.c...
★ Reference: Time series analysis, James D. Hamilton Princeton University Press, 1994.