A Model of Motion Processing in the Visual Cortex Using Neural Field With Asymmetric Hebbian Learning

Abstract

Neurons in the dorsal pathway of the visual cortex are thought to be involved in motion processing. The first site of motion processing is the primary visual cortex (V1), encoding the direction of motion in local receptive fields, with higher order motion processing happening in the middle temporal area (MT). Complex motion properties like optic flow are processed in higher cortical areas of the Medial Superior Temporal area (MST). In this study, a hierarchical neural field network model of motion processing is presented. The model architecture has an input layer followed by either one or cascade of two neural fields (NF): the first of these, NF1, represents V1, while the second, NF2, represents MT. A special feature of the model is that lateral connections used in the neural fields are trained by asymmetric Hebbian learning, imparting to the neural field the ability to process sequential information in motion stimuli. The model was trained using various traditional moving patterns such as bars, squares, gratings, plaids, and random dot stimulus. In the case of bar stimuli, the model had only a single NF, the neurons of which developed a direction map of the moving bar stimuli. Training a network with two NFs on moving square and moving plaids stimuli, we show that, while the neurons in NF1 respond to the direction of the component (such as gratings and edges) motion, the neurons in NF2 (analogous to MT) responding to the direction of the pattern (plaids, square object) motion. In the third study, a network with 2 NFs was simulated using random dot stimuli (RDS) with translational motion, and show that the NF2 neurons can encode the direction of the concurrent dot motion (also called translational flow motion), independent of the dot configuration. This translational RDS flow motion is decoded by a simple perceptron network (a layer above NF2) with an accuracy of 100% on train set and 90% on the test set, thereby demonstrating that the proposed network can generalize to new dot configurations. Also, the response properties of the model on different input stimuli closely resembled many of the known features of the neurons found in electrophysiological studies.

A biologically plausible dynamic deep network for recognizing structure from motion and biological motion

“…The model is an expanded version of an earlier model (Gundavarapu et al, 2019). In overview, the stage-1 VSMN generates velocity-sensitive representations of moving dots.…”

“…We present a dynamic deep network model that can (i) recognize 3D shape of a rotating pattern (ii) identify an action from a PL display, under conditions of challenging noise backgrounds. The model is an expanded version of an earlier model (Gundavarapu et al, 2019). In overview, the stage-1 VSMN generates velocity-sensitive representations of moving dots.…”

“…We test the proposed model with both rotating shapes and biological point-light displays, and drawn conclusions based on the network performance. Earlier we presented a biologically plausible computational model (Gundavarapu et al, 2019) that can explain neural responses at V1 and MT to motion stimuli. This dynamic model was recently extended using a deep convolutional neural network (CNNs) for optic flow recognition and was described in a companion paper (Gundavarapu and Chakravarthy, 2022).…”

The development of bi-directionally coupled self-organizing neurovascular networks captures orientation-selective neural and hemodynamic cortical responses

“…The lateral connections with short-range excitatory projections and long-range inhibitory projections facilitate competitive learning and are ideal for modeling the retinotopic map formation seen in the primary visual cortex (V1). The lateral connections trained using asymmetric Hebbian learning [31] make the neural units naturally direction sensitive. There are advanced models of LISSOM like Gain-Controlled Adaptive LISSOM (GCAL) [54,55],…”

“…A LISSOM used to simulate the neural layer is well known to have a lateral connectivity pattern with near excitatory (ON center) and far inhibitory (OFF surround) connections [27,31]. But this is not the case when the network is used to simulate the vascular layer.…”