- Location: Wilson Hall • 111 21St Ave S • Nashville, TN 37240
- Room: 115
- Contact: Angel Gaither
- Email: email@example.com
- Phone: 615-322-0080
- Website: https://www.vanderbilt.edu/psychological_sciences/events/index.php
- Audience: Free and Open to the Public
Oakyoon Cha, PhD
Department of Psychology (Blake & Gauthier Labs)
"Ensemble as Lossy Compression Rather Than Dumb Averaging"
Our ability to represent ensembles, i.e., statistical properties, of visual information has been suggested as a building block of gist perception. To effectively summarize different groups of objects into a scene gist, ensembles should be more than the average of visual features in a scene. In one study, I found that visual features belonging to different perceptual groups (surface and contour) are not likely to blend into a single ensemble representation. This tendency may be attributed to people’s experiences in keeping the visual features of occluding objects separate from those of the occluded objects. In another study, I explored people’s ability to summarize complex visual stimuli that cannot be mapped onto a linear or circular dimension, such as a group of cars. Indeed, people learned to determine the diversity of a group of cars, and their strategies depend on their experiences with repeated stimulus features. These two studies suggest that ensemble computations are based on complex, task-relevant information rather than simple, task-agnostic averaging.