Abstract
The development of the retinotopic map is presumed to precede the development of the orientation map in V1 in primates. Experimental studies demonstrate a radial bias, wherein radial orientations produce higher activity compared to other orientations [1]. However most traditional models assume isometry in the orientation map developed, i.e. the orientation maps throughout V1 are similar in nature, independent of its retinotopy. In this paper, we propose an activity-dependent model which simulates the development of a radially biased orientation map. To that end we simulate the large-scale development of the retinotopic map, followed by the development of the orientation map in a sub-region of this map. The architecture consists of a Laterally Interconnected Synergetically Self Organizing Map (LISSOM) [2] with 2 layers, representing the retina, and the V1 respectively (see Figure Figure1A).1A). At each time step, each neuron in V1, combines the afferent activation (ζr1,r2) along with its lateral excitations and inhibitions (ηkl) from the previous time step. Figure 1 (A) Schematic representation of the LISSOM architecture. (B) V1 orientation map developed after initial training to establish retinotopy along with its color map. (C) Orientation sub-maps, biased by initial retinotopy, at 1000, 2000, 3000, 4000 iterations. ... ηij(t)=σ(∑r1,r2ζr1,r2 μij,r1r2+ γE∑k,lEij,kl ηkl(t-1)-γI∑k,lIij,kl ηkl(t-1)) (1) The afferent (μij,r1r2), lateral excitatory (Eij,kl) and lateral inhibitory (Iij,kl) weights adapt based on a normalized Hebbian mechanism. In order to develop the retinotopic map, the inputs to the retinal layer consists of centered (assumed to be the point of fixation) rectangular bars of varying dilations and rotations as modelled in [3]. The retinotopic map developed, biases the initial configuration of the orientation map (see Figure Figure1B)1B) since all the bars given during the initial training are centered. For the subsequent refinement of the orientation map, Gaussians of differing orientation and positions (non-centered) are given as inputs to the retinal layer. After training for 4000 iterations (see Figure 1C,D), it is observed that the developed orientation map prefers those orientations which the retinotopy biases it towards, quantified by their corresponding histograms (See Figure 1E,F). As seen from the histogram the area occupied by the region mapping 1250-1500 is larger in the map developed assuming retinotopic bias, compared to that of the map developed assuming isotropy.
Highlights
The development of the retinotopic map is presumed to precede the development of the orientation map in V1 in primates
The afferent, lateral excitatory (Eij,kl) and lateral inhibitory (Iij,kl) weights adapt based on a normalized Hebbian mechanism
The retinotopic map developed, biases the initial configuration of the orientation map since all the bars given during the initial training are
Summary
The development of the retinotopic map is presumed to precede the development of the orientation map in V1 in primates. Ηij(t) = σ ( ζr1,r2 μij,r1r2 + γE Eij,kl ηkl(t − 1)− γI Iij,kl ηkl(t − 1))
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.