[Do not cite without permission]
BibTeX
@MISC{Huber_[donot,
author = {David E. Huber and David E. Huber},
title = {[Do not cite without permission]},
year = {}
}
OpenURL
Abstract
This chapter is presented in three sections corresponding to three models that incorporate Bayesian explaining away between different sources. The first section considers primes and targets as potential sources without reference to time. The original ROUSE model is reformulated as a generative model, arriving at the original equations but with slightly different dependence assumptions. The second section considers a model in which past time steps explain away future times steps, thereby producing perceptual sensitivity to the onset of new objects (i.e., new events). The resultant dynamics are related to the dynamics of neural habituation in several important ways. The third section considers a model in which future time steps explain away past time steps, thereby producing sensitivity to the offset of old objects (i.e., old events). By cascading layers, a working memory system is developed that represents the temporal rank ordering of objects regardless of their specific durations (i.e., scale free sequential information).Causality in Time 3 In recent years, Bayesian models of cognition have effectively explained a wide variety of cognitive behaviors ranging from visual perception (Yuille & Kersten, 2006)







