Recurrent network dynamics reconciles visual motion segmentation and integration

Abstract

In sensory systems, a range of computational rules are presumed to be implemented by neuronal subpopulations with different tuning functions. For instance, in primate cortical area MT, different classes of direction-selective cells have been identified and related either to motion integration, segmentation or transparency. Still, how such different tuning properties are constructed is unclear. The dominant theoretical viewpoint based on a linear-nonlinear feed-forward cascade does not account for their complex temporal dynamics and their versatility when facing different input statistics. Here, we demonstrate that a recurrent network model of visual motion processing can reconcile these different properties. Using a ring network, we show how excitatory and inhibitory interactions can implement different computational rules such as vector averaging, winner-take-all or superposition. The model also captures ordered temporal transitions between these behaviors. In particular, depending on the inhibition regime the network can switch from motion integration to segmentation, thus being able to compute either a single pattern motion or to superpose multiple inputs as in motion transparency. We thus demonstrate that recurrent architectures can adaptively give rise to different cortical computational regimes depending upon the input statistics, from sensory flow integration to segmentation.

Center-surround velocity-based segmentation: Speed, eccentricity, and timing of visual stimuli interact to determine interocular dominance

“…If both stimuli were fast, both movement directions were visible simultaneously with the appearance of semitransparency (Blake et al, 1985;Marshak & Sekuler, 1979;van de Grind et al, 2001). This result is consistent with the modeling work suggesting that higher divergence in motion signal leads to transparency (Medathati, Rankin, Meso, Kornprobst, & Masson, 2017). In terms of modeling the velocity response, higher velocity divergence between the two rivaling motion clouds should result in more defined velocity peaks.…”

“…In terms of modeling the velocity response, higher velocity divergence between the two rivaling motion clouds should result in more defined velocity peaks. Medathati et al (2017) show that, when the two motion signals cannot be reconciled (vector averaging) or result in a clear dominance (winner takes all), the two motion carriers are seen simultaneously as spatially overlapping, resulting in transparency.…”

“…The motion cloud configuration used here corresponds to an intermediate motion variance between pure gratings (single spatial frequency) and random dot patterns (broadband stimulus). Medathati et al's (2017) work, thus, suggests that spatial frequency range is important, and this should be addressed in the future.…”

“…However, an important future direction is to determine exactly which parameters lead to the departure from the rivalry observed with the popular random dot kinematograms (Blake et al, 1985;Marshak & Sekuler, 1979;van de Grind et al, 2001;Wiesenfelder & Blake, 1990). Computational work suggests that we should expect differences in dominance patterns if stimulus parameters are varied (Medathati et al, 2017). The motion cloud configuration used here corresponds to an intermediate motion variance between pure gratings (single spatial frequency) and random dot patterns (broadband stimulus).…”

Criterion-free measurement of motion transparency perception at different speeds

“…Constituent directions separated by at least 60° elicited bimodal responses in pattern cells, consistent with previous work (Qian et al, 1994 ; Treue, Hol, & Rauber, 2000 ). This strongly implicates the circuits that generate pattern selectivity in transparent motion perception (but see Xiao & Huang, 2015 for a different interpretation)—a long-standing prediction of Simoncelli and Heeger's ( 1998 ) model of motion processing in MT (but see Medathati, Rankin, Meso, Kornprobst, & Masson, 2017 for an alternative computational account).…”