Adaptive Time Series Analysis Using Predictive Inference and Entropy.
Abstract
Research is being conducted on adaptive time series methods for detecting and tracking both abrupt and slow changes in both structure and parameters. The methods are based on a unified statistical frame work which is motivated by statistical inference and entropy arguments. The method yields estimates of input/output dynamics and noise statistics. An integrated approach which combines canonical variates analysis and maximum likelihood estimation has been developed and tested. Specific attention is given to the problem of parameter truncation in both a linear predictor and Kalman filter framework.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 1987
- Accession Number
- ADA191858
Entities
People
- Donald E. Dustafson