Show Reference: "Ideal-Observer Models of Cue Integration"

Ideal-Observer Models of Cue Integration In Sensory Cue Integration (August 2011), pp. 251-262 by Michael S. Landy, Martin S. Banks, David C. Knill edited by Julia Trommershäuser, Konrad Körding, Michael S. Landy
    address = {Oxford},
    author = {Landy, Michael S. and Banks, Martin S. and Knill, David C.},
    booktitle = {Sensory Cue Integration},
    editor = {Trommersh\"{a}user, Julia and K\"{o}rding, Konrad and Landy, Michael S.},
    keywords = {bayes, multi-modality},
    month = aug,
    pages = {251--262},
    posted-at = {2011-12-14 14:31:55},
    priority = {2},
    publisher = {Oxford University Press},
    title = {{Ideal-Observer} Models of Cue Integration},
    url = {\~{}david/courses/perceptionGrad/Readings/LandyBanksKnill-2011.pdf},
    year = {2011}

See the CiteULike entry for more info, PDF links, BibTex etc.

MLE has been a successful model in many sensory cue integration tasks.

According to Landy et al., humans often combine cues (intra- or cross-sensory) optimally, consistent with MLE.

Landy et al. specifically state that one conclusion one can draw from observing sub-optimal behavior in a biological system is that the task may be too different from the natural tasks doing which the system was shaped.

We cannot always expect optimal behavior in tasks which have become relevant only recently in human development, like eg. in complex reasoning tasks, or in tasks with highly artificial stimuli.

When studying an information-processing system, and given a computational theory of it, algorithms and representations for implementing it can be designed, and their performance can be compared to that of natural processing.

If the performance is similar, that supports our computational theory.

Nature has had millions of years to optimize the performance of cognitive systems. It is therefore reasonable to assume that they perform optimally wrt. natural tasks and natural conditions.

Bayesian theory provides a framework to determine optimal strategies. Therefore, it makes sense to operate under the assumption that the processes we observe in nature can be understood as implementations of Bayes-optimal strategies.