Petkov N, 2004, "Framework for testing biological-utility hypotheses" Perception 33 ECVP Abstract Supplement
Framework for testing biological-utility hypotheses
Various aspects of the visual system have been claimed to have certain biological role. The usual argument is that a given neural mechanism would help an animal to perform better a certain task. Such speculations are typically not supported by quantitative measurements and statistical studies. I propose a framework for verification of biological-utility hypotheses that includes defining a measure that quantifies the performance of a given subsystem in a certain task and collecting statistics about this measure on a reasonable number of images. A hypothesis is confirmed or rejected by using the distribution of the performance improvement/degradation values obtained by taking the considered mechanism into account in a computational model of the concerned subsystem. This framework for utility verification follows an approach previously deployed for the comparison of biologically motivated contour-detection algorithms in a computer science context (Grigorescu et al, 2003 IEEE Transactions on Image Processing 12 729 - 739). The framework is applied to non-classical receptive-field inhibition for which the detection of object contours (Petkov and Westenberg, 2003 Biological Cybernetics 88 236 - 246) and texture boundaries (Nothdurft et al, 1999 Visual Neuroscience 16 15 - 34) has been suggested as probable biological utility.
These web-based abstracts are provided for ease of seaching and access, but certain aspects (such as as mathematics) may not appear in their optimum form. For the final published version of this abstract, please see
ECVP 2004 Abstract Supplement (complete) size: 1686 Kb