The Stabilized Supralinear Network: A Unifying Circuit Motif Underlying Multi-Input Integration in Sensory Cortex

Abstract

Summary Neurons in sensory cortex integrate multiple influences to parse objects and support perception. Across multiple cortical areas, integration is characterized by two neuronal response properties: (1) surround suppression: modulatory contextual stimuli suppress responses to driving stimuli; (2) “normalization”: responses to multiple driving stimuli add sublinearly. These properties depend on input strength: for weak driving stimuli, contextual influences more weakly suppress or facilitate and summation becomes linear or supralinear. Understanding the circuit operations underlying integration is critical to understanding cortical function and disease. We present a simple, general theory. A wealth of integrative properties including the above emerge robustly from four properties of cortical circuitry: (1) supralinear neuronal input/output functions; (2) sufficiently strong recurrent excitation; (3) feedback inhibition; (4) simple spatial properties of intracortical connections. Integrative properties emerge dynamically as circuit properties, with excitatory and inhibitory neurons showing similar behaviors. In new recordings in visual cortex, we confirm key model predictions.

Balanced feedforward inhibition and dominant recurrent inhibition in olfactory cortex

“…Higher stimulation intensities likely recruit intracortical excitation and, ultimately, dominant L2/3 recurrent inhibition. These circuit features are consistent with balanced and inhibition stabilized network models (12,13,15) that replicate supralinear and sublinear computations in visual and auditory cortex (5,7). Thus, feedforward and recurrent inhibition in the piriform cortex could represent global mechanisms for sensory processing throughout the brain.…”

“…In the latter case, odors evoke strong, broadly tuned intracortical excitation (4). The recruitment of strong feedback inhibition under these conditions could function to prevent runaway excitation and stabilize the network, resulting in sublinear or normalization of cortical responses (15,45). Consistent with this idea, disinhibition of the piriform cortex leads to epileptic activity (46)(47)(48)(49).…”

“…In sensory cortices, including the olfactory cortex, neural responses to sensory stimuli depend on the relative balance of inhibition with respect to excitation in both feedforward and recurrent pathways (2-7). Moreover, numerous theoretical studies have suggested that balanced cortical networks underlie the selectivity, sparseness, and correlations of cortical activity (8)(9)(10)(11)(12)(13)(14)(15). These studies highlight the importance of quantifying the relationship between excitation and inhibition in cortical networks.…”

Eigenspectrum bounds for semirandom matrices with modular and spatial structure for neural networks

“…Our results are suggestive of stability and operating regimes also for networks with nonlinear transfer functions, but the shape of the neuron gain function can introduce further complexity into the operating modes of a network [42,43].…”

Suppressive mechanisms in visual motion processing: From perception to intelligence

“…The answer to this question is important not only for our understanding of motion perception, but also for elucidation of basic principles that underlie visual processing in general. While various types of visual cues can help guide integration and segregation of motion signals (Rivest & Cavanagh, 1996; Croner & Albright, 1997, 1999; Lorenceau & Alais, 2001; Tadin et al, 2002), the quality of sensory signals per se is an important factor determining the appropriate balance between integrating and differentiating processes (Faisal, Selen & Wolpert, 2008; Rubin, Van Hooser & Miller, 2015). Research by the author and other groups (Section 3.2) shows that the nature of spatial integration of motion adapts to visual conditions, with spatial summation giving way to spatial suppression as stimulus saliency increases.…”