Deutsch
 
Hilfe Datenschutzhinweis Impressum
  DetailsucheBrowse

Datensatz

DATENSATZ AKTIONENEXPORT

Freigegeben

Poster

Eye movements for active learning of objects

MPG-Autoren
/persons/resource/persons84248

Tanner,  TG
Department Human Perception, Cognition and Action, Max Planck Institute for Biological Cybernetics, Max Planck Society;
Max Planck Institute for Biological Cybernetics, Max Planck Society;

/persons/resource/persons83913

Fleming,  R
Research Group Computational Vision and Neuroscience, Max Planck Institute for Biological Cybernetics, Max Planck Society;
Department Human Perception, Cognition and Action, Max Planck Institute for Biological Cybernetics, Max Planck Society;
Max Planck Institute for Biological Cybernetics, Max Planck Society;

/persons/resource/persons83839

Bülthoff,  HH
Department Human Perception, Cognition and Action, Max Planck Institute for Biological Cybernetics, Max Planck Society;
Max Planck Institute for Biological Cybernetics, Max Planck Society;

Volltexte (beschränkter Zugriff)
Für Ihren IP-Bereich sind aktuell keine Volltexte freigegeben.
Volltexte (frei zugänglich)
Es sind keine frei zugänglichen Volltexte in PuRe verfügbar
Ergänzendes Material (frei zugänglich)
Es sind keine frei zugänglichen Ergänzenden Materialien verfügbar
Zitation

Tanner, T., Fleming, R., & Bülthoff, H. (2007). Eye movements for active learning of objects. Poster presented at 7th Annual Meeting of the Vision Sciences Society (VSS 2007), Sarasota, FL, USA.


Zitierlink: https://hdl.handle.net/11858/00-001M-0000-0013-CD79-B
Zusammenfassung
We investigated how humans use eye movements to direct their attention to informative features in a categorization task. More specifically, we test the hypothesis that eye movements are influenced by prior knowledge about a task and by information gathered in previous fixations. Our novel stimuli, which belonged to either one of two probabilistic classes, were large circular contours with several regular perturbations at which the curvature was varied as a continous feature dimension. With this design the spatial separation of single features generally required several closer fixations to make a confident decision about class membership.

Each feature value varied stochastically from trial to trial according to a characteristic distribution for each category (external noise). The features were independent and varied in diagnosticity. Subjects had to learn the categories by using immediate feedback about the true category after each trial (4 subjects, 10 sessions of 250 trials). We estimated the internal noise, which was much smaller then the external noise, based on an independent experiment measuring curvature discrimination performance for different eccentricites (0–12°), finding approx. linear decrease in sensitivity with increasing curvature.

The subjects were able to learn to discriminate the categories (avg. performance for ideal observer vs. subjects was 0.82 vs 0.68). Trial by trial fluctations in performance follow the ideal observer (MAE 0.32). With increasing expertise reaction times became shorter and fixations became more focused, possibly reflecting the subjects' belief about relevant features. We compare the results with Bayesian learner models which take into account the peripheral fall-off in discriminability, while directing their attention to the currently most informative features.