General Multidecision Theory: Hypothesis Testing and Changepoint Detection with Applications to Homeland Security.

摘要

The objective was to develop general theories of sequential hypothesis testing and quickest change detection for complex multi-population stochastic models, as well as to apply these theories to automatic threat detection and classification with low false alarm and miss-classification rates. More specifically, we addressed complex stochastic models that include: multi-population/multi-channel models; multi-hypothesis scenarios; general stochastic models with non-stationary and dependent observations; prior uncertainty. Within a limited duration of the project (about 8 months) a background for a general theory of testing multiple composite hypotheses was established, and certain particular examples were considered.