Stopping Criterion Design for Recursive Bayesian Classification: Analysis and Decision Geometry

  • 2020-07-30 16:21:10
  • Aziz Kocanaogullari, Murat Akcakaya, Deniz Erdogmus
  • 1


Systems that are based on recursive Bayesian updates for classification limitthe cost of evidence collection through certain stopping/termination criteriaand accordingly enforce decision making. Conventionally, two terminationcriteria based on pre-defined thresholds over (i) the maximum of the stateposterior distribution; and (ii) the state posterior uncertainty are commonlyused. In this paper, we propose a geometric interpretation over the stateposterior progression and accordingly we provide a point-by-point analysis overthe disadvantages of using such conventional termination criteria. For example,through the proposed geometric interpretation we show that confidencethresholds defined over maximum of the state posteriors suffer from stiffnessthat results in unnecessary evidence collection whereas uncertainty basedthresholding methods are fragile to number of categories and terminateprematurely if some state candidates are already discovered to be unfavorable.Moreover, both types of termination methods neglect the evolution of posteriorupdates. We then propose a new stopping/termination criterion with ageometrical insight to overcome the limitations of these conventional methodsand provide a comparison in terms of decision accuracy and speed. We validateour claims using simulations and using real experimental data obtained througha brain computer interfaced typing system.


Quick Read (beta)

loading the full paper ...