crta
Hrvatska znanstvena Sekcija img
bibliografija
3 gif
 Naslovna
 O projektu
 FAQ
 Kontakt
4 gif
Pregledavanje radova
Jednostavno pretraživanje
Napredno pretraživanje
Skupni podaci
Upis novih radova
Upute
Ispravci prijavljenih radova
Ostale bibliografije
Slični projekti
 Bibliografske baze podataka

Pregled bibliografske jedinice broj: 407764

Zbornik radova

Autori: Šetić, Mia; Domijan, Dražen
Naslov: A Neural Model of Spatial Shifts of Attention
Izvornik: Twelfth International Confertence on Cognitve and Neural Systems - ProceedingsBoston :
Skup: Twelfth International Conference on Cognitve and Neural Systems
Mjesto i datum: Boston, SAD, 14.-17. 05. 2008.
Ključne riječi: Attention; Visual Search; Neural Model
Sažetak:
Computational models of visual search assume the existence of a saliency map, which represents the measure of the relative importance of spatial locations in the visual field. Points of interest are read-out from the saliency map by lateral inhibition. In order to move attention to a new location, the most active location is inhibited, which allows the selection of the next most salient location. This is known as inhibition of return (Itti & Koch, 2001). Important constraint of the proposed mechanism for inhibition of return is that it is not consistent with several psychophysical findings about visual search (Horowitz et al., 2004 ; Ogawa et al., 2002 ; Pashler, et al., 2004). We proposed a new computational model for the inhibition of return, which is able to examine saliency map in a manner consistent with recent psychophysical findings. The basic elements of the model are two neural integrators connected to two inhibitory interneurons. The integrators keep the saliency value of the currently attended location in the working memory. The inhibitory interneurons modulate a feedforward flow of information from the saliency map to the output map, which points to the location of interest. The neural integrators together with inhibitory interneurons provide non-linear gating of input from saliency map. Activity of the neural integrators is controlled by the match of the top-down target pattern and bottom-up or input pattern loaded at the location of interest (Grossberg et al., 1994). Mismatch between patterns creates a change in the relative activity levels of neural integrators. First, non-linear gate is closed by reducing the activity difference between integrators, which removes currently loaded input pattern. After that, non-linear gate is open again by increase in activity difference between integrators, which initiates search for another input pattern. Exchange between opening and closing of non-linear gate continues until match between target and input pattern is found or until all locations are visited. Computer simulations illustrate that the proposed model can select multiple locations simultaneously since it does not depend on the recurrent lateral inhibition (Grossberg et al., 1994). The model is able to search the saliency map even if objects move randomly in visual space (Ogawa et al., 2002). Moreover, the model exhibits different dynamics for spatial shifts of attention and attentional tracking (Horowitz et al., 2004). Also, the model has ability to search among low contrast stimuli (Pashler et al., 2004). Finally, the model is a simple feedforward network without recurrent connections, which enable it to perform fast searches.
Vrsta sudjelovanja: Poster
Vrsta prezentacije u zborniku: Sažetak
Vrsta recenzije: Međunarodna recenzija
Projekt / tema: 009-0362214-0818
Izvorni jezik: ENG
Kategorija: Znanstveni
Znanstvena područja:
Psihologija
Upisao u CROSBI: ddomijan@ffri.hr (ddomijan@ffri.hr), 8. Lip. 2009. u 15:20 sati



Verzija za printanje   za tiskati


upomoc
foot_4