EM algorithm convergence KF log likelihood decrease

Click For Summary
SUMMARY

The discussion centers on the convergence of the Expectation-Maximization (EM) algorithm in relation to the Kalman Filter (KF) and the observed decrease in log likelihood for the model parameters p(Y_{k}|Y_{k-1}). Users suggest that a decrease in log likelihood may indicate issues with the implementation of the EM algorithm. Additionally, it is recommended to explore existing code repositories, particularly on the R platform, to compare results and improve the implementation.

PREREQUISITES
  • Understanding of the Expectation-Maximization (EM) algorithm
  • Familiarity with Kalman Filter (KF) techniques
  • Knowledge of log likelihood in statistical modeling
  • Experience with R programming for statistical analysis
NEXT STEPS
  • Examine existing R packages that implement the EM algorithm for parameter estimation
  • Research techniques for improving log likelihood convergence in Kalman Filters
  • Learn about diagnostic tools for evaluating EM algorithm performance
  • Explore alternative statistical methods for parameter fitting beyond EM
USEFUL FOR

Data scientists, statisticians, and machine learning practitioners who are implementing the EM algorithm and Kalman Filters for parameter estimation and model optimization.

MikeLowri123
Messages
10
Reaction score
0
Hi everyone,

Im running the KF to learn parameters of a model, the log likelihood of the p(Y_{k}|Y_{k-1}), however decreases.

Can anyone advise, does this mean my implementation is wrong or can this just be the case.

Advice appreciated

Thanks
 
Physics news on Phys.org
Can anyone offer a small piece of advice or even a reference??
 
Hey MikeLowri123.

If you are using the EM to fit some parameters for a parametric distribution, have you tried grabbing some code from a repository like the R platform and seeing the results?
 

Similar threads

  • · Replies 6 ·
Replies
6
Views
3K
  • · Replies 0 ·
Replies
0
Views
1K
  • · Replies 3 ·
Replies
3
Views
2K
  • · Replies 12 ·
Replies
12
Views
3K
  • · Replies 3 ·
Replies
3
Views
2K
  • · Replies 7 ·
Replies
7
Views
2K
  • · Replies 13 ·
Replies
13
Views
4K
  • · Replies 9 ·
Replies
9
Views
2K
  • · Replies 62 ·
3
Replies
62
Views
12K
Replies
4
Views
3K