Single circuit in V1 capable of switching contexts during movement using VIP population as a switch

Abstract

As animals adapt to their environments, their brains are tasked with processing stimuli in different sensory contexts. Whether these computations are context dependent or independent, they are all implemented in the same neural tissue. A crucial question is what neural architectures can respond flexibly to a range of stimulus conditions and switch between them. This is a particular case of flexible architecture that permits multiple related computations within a single circuit.Here, we address this question in the specific case of the visual system circuitry, focusing on context integration, defined as the integration of feedforward and surround information across visual space. We show that a biologically inspired microcircuit with multiple inhibitory cell types can switch between visual processing of the static context and the moving context. In our model, the VIP population acts as the switch and modulates the visual circuit through a disinhibitory motif. Moreover, the VIP population is efficient, requiring only a relatively small number of neurons to switch contexts. This circuit eliminates noise in videos by using appropriate lateral connections for contextual spatio-temporal surround modulation, having superior denoising performance compared to circuits where only one context is learned. Our findings shed light on a minimally complex architecture that is capable of switching between two naturalistic contexts using few switching units.Author SummaryThe brain processes information at all times and much of that information is context-dependent. The visual system presents an important example: processing is ongoing, but the context changes dramatically when an animal is still vs. running. How is context-dependent information processing achieved? We take inspiration from recent neurophysiology studies on the role of distinct cell types in primary visual cortex (V1).We find that relatively few “switching units” — akin to the VIP neuron type in V1 in that they turn on and off in the running vs. still context and have connections to and from the main population — is sufficient to drive context dependent image processing. We demonstrate this in a model of feature integration, and in a test of image denoising. The underlying circuit architecture illustrates a concrete computational role for the multiple cell types under increasing study across the brain, and may inspire more flexible neurally inspired computing architectures.

A biologically inspired architecture with switching units can learn to generalize across backgrounds

“…This strategy overcomes catastrophic forgetting if the network can solve the task on dataset 2 by only learning the weights from the switching units. As expected from computational work studying the complex recurrence of the V1 circuit [55], we find that this feedforward switching network achieves accuracies on dataset 2 (Figure 3C, green line) only slightly higher than the unmatched condition when using a network trained on dataset 1 alone (orange line), and far below the reference matched condition, when the network was trained on dataset 2 (blue line).…”

“…A third, related strategy takes inspiration from the biological V1 circuit discussed above [55,65], where VIP neurons turn ON and OFF depending on an animal's motion context. In our first, simplest model inspired by this circuit, we add switching units akin to the VIP neurons: they are OFF for the first context and ON for the second context.…”

“…We can improve the feedforward switching network by adding recurrent connections to the switching units (blue arrows, Figure 3A) [55]. When adding switching units to convolutional layers, we essentially add kernels, and we refer to switching units per unit space (sw.u/u.s.).…”

“…When an animal shifts from stationary to moving behavior, a population of neurons in the primary visual cortex (V1) which belongs to a cell class expressing Vasoactive Intestinal Peptide -the VIP population -increases its activity. It has been hypothesized that this neuron population turns ON like a switch and reconfigures the circuit dynamics to efficiently process the corresponding static vs. moving visual scenes [55,65]. Following this, we propose a network architecture, the switching network, where we add units that are OFF when the network is presented the first context, but turn ON for the second context, similarly to the VIP (Figure 1F,G).…”