Statistics Seminars: NPI: An application to classification
7 December 2009 14:15 in CM221
In this presentation, we show how the NPI model for multinomial data can be used to build classification trees. When using imprecise models in the construction of classification trees, it is necessary to find the maximum entropy probability distribution consistent with the model. We present two algorithms, one approximate and one exact, for finding the maximum entropy distribution consistent with NPI. We then compare our methods with previous approaches including precise and imprecise methods.
Contact firstname.lastname@example.org for more information