Predictions of a model of spatial attention using sum- and max-pooling functions

Abstract

Assuming a convergent projection within a hierarchy of processing stages stimuli from different areas of the receptive ÿeld project onto the same population of cells. Pooling over space a ects the representation of individual stimuli, and thus its understanding is crucial for attention and ultimately for object recognition. Since attention, in turn, is likely to modify such spatial pooling by changing the competitive weight of individual stimuli, we compare the predictions of sum-and max-pooling methods using a model of attention. Both pooling functions can account for data investigating the competition between a pair of stimuli within a V4 receptive ÿeld; however, our model using sum-pooling predicts a di erent tuning curve. If we present an additional probe stimulus with the pair, sum-pooling predicts a bottom-up bias of attention, whereas the competition for attention using max-pooling is robust against the additional stimulus.

Split of spatial attention as predicted by a systems‐level model of visual attention

“…However, in the following studies they are strongly inhibited by fixation cells and thus are inactive. Modulatory spatial feedback into V4 (Hamker, 2004b) now originates in FEFvm cells and not FEFm cells as in previous versions.…”

Attention selectively gates afferent signal transmission to area V4

“…This does not imply a change of spike rates in the neuronal population delivering these afferents. Such selective input gating has been proposed by several models implementing biased competition (8)(9)(10) and may implement multiplication of input signals assumed by divisive normalization models of attention (24,28,29). Furthermore, a modulation of the efficacy of inputs to V4 neurons has been suggested to explain the lack of latency effects despite of average firing rate increases with attention (30).…”

“…As a result of such selective input gating, local neuronal processing would be predominately determined by signals representing the attended stimulus (e.g. 8,[9][10][11]. The second class assumes that attention could act directly on the response strength of the neurons.…”

Modeling Attention: From Computational Neuroscience to Computer Vision

“…It has been suggested that the gain of a neuronal response to excitatory drive is decreased by increasing the level of both, excitatory and inhibitory, background firing rates in a balanced manner [1]. On a more abstract level a feedback signal could increase the gain of the feedforward pathway in a multiplicative fashion [18], [9], [22]. We investigated such a gain control mechanism by simulating a V4 layer which receives input from a V2 population.…”