Geometric interpretation of robustness in spike coding networks

Models of face, object, and scene recognition traditionally focus on massively parallel processing of low- level features, with higher order representations emerging at later processing stages [1,2,3]. However, visual perception relies on eye movements, which are necessarily sequential. Neurons in entorhinal cortex with grid cell-like firing have recently been reported in response to eye movements, i.e. in visual space [4,5,6]. A functional explanation for these ‘visual grid cells’ is so far lacking. Grid cells (GCs) [7] are predominantly known from spatial memory and navigation studies and exhibit regularly arranged firing fields in the navigation plan. Ensembles of GCs are thought to underlie path integration and the calculation of goal- directed movement vectors [8,9,10]. Following the presumed role of GCs in vector navigation, we propose a model of recognition memory for familiar faces, objects, and scenes, in which GCs encode translation vectors between salient stimulus features. A sequence of saccadic eye movement vectors, moving from one salient feature to the expected location of the next, potentially confirms an initial hypothesis (accumulating evidence towards a threshold) about stimulus identity. This identification is based on the relative feature layout (going beyond recognition of individual features), and implements an relational active sensing strategy to infer the stimulus identity of exemplars within a stimulus category. Category identification is hypothesised to occur in earlier occipito-temporal areas, likely via parallel processing. The model constitutes the first quantitative proposal for a role of GCs in visual recognition. The variance of GC activity along saccade trajectories exhibits 6-fold symmetry across 360 degrees akin to recently reported fMRI data [5,6]. The mechanism is robust with regard to partial visual occlusion, can accommodate size and position invariance, and suggests a functional explanation for medial temporal lobe involvement in visual memory for relational information and memory guided attention