Mouse visual cortex areas represent perceptual and semantic features of learned visual categories

Abstract

Associative memories are stored in distributed networks extending across multiple brain regions. However, it is unclear to what extent sensory cortical areas are part of these networks. Using a paradigm for visual category learning in mice, we investigated whether perceptual and semantic features of learned category associations are already represented at the first stages of visual information processing in the neocortex. Mice learned categorizing visual stimuli, discriminating between categories and generalizing within categories. Inactivation experiments showed that categorization performance was contingent on neuronal activity in the visual cortex. Long-term calcium imaging in nine areas of the visual cortex identified changes in feature tuning and category tuning that occurred during this learning process, most prominently in the postrhinal area (POR). These results provide evidence for the view that associative memories form a brain-wide distributed network, with learning in early stages shaping perceptual representations and supporting semantic content downstream.

Amplified cortical neural responses as animals learn to use novel activity patterns

“…Previous studies suggest that learning in visual perceptual tasks can lead to changes in the tuning properties of responsive neurons in mouse V1 (Goltstein et al, 2021; Khan et al, 2018). However, it remains unresolved if these changes arise from plasticity in the local cortical networks or if changes may be inherited from thalamic input pathways that could in principle adjust input strength, state, or synchrony (Cano et al, 2006; Hubel and Wiesel, 1962; Kelly et al, 2014; Sadagopan and Ferster, 2012) to change cortical responses.…”

“…There is substantial evidence that sensory cortical synapses can be modified by activity (Frégnac and Shulz, 1999;He et al, 2006;Hengen et al, 2013;Malenka and Bear, 2004;Sawtell et al, 2003), but it is less clear whether cortical response changes constitute the computational change that leads to improved behavior with learning. Studies in humans and animals have reported varied effects of learning on visual cortical responses, including increased activity after visual training (Bao et al, 2010;Li et al, 2008;Schoups et al, 2001;Schwartz et al, 2002), selective suppression of activity (Ghose et al, 2002), decreased variability of visual selectivity response properties after training (Goltstein et al, 2013(Goltstein et al, , 2021Poort et al, 2015), and activity changes that disappeared once early learning has ended (Yotsumoto et al, 2008). Some learning studies have found improvement in primary sensory representations (Goltstein et al, 2021;Henschke et al, 2020;Jurjut et al, 2017;Marshel et al, 2019), along with changes in anticipatory and other signals (Khan et al, 2018;Poort et al, 2015).…”

Population codes enable learning from few examples by shaping inductive bias

“…More work is need to study other factors that affect inductive bias. Importantly, sensory neuron tuning curves can adapt during perceptual learning tasks [86, 87, 88, 89] with the strength of adaptation dependent on brain area [90, 91, 92, 93]. This motivates analysis of learning in multi-layer networks.…”

Are Grid-Like Representations a Component of All Perception and Cognition?

“…If the frontal cortex is engaged in representing knowledge in a generalized cognitive map, it is not completely impossible that animals use a similar principle as shown in human fMRI findings. This prediction may be tested in freely foraging rodents and monkeys, as animals have demonstrated the capability of learning the category knowledge (Fize et al, 2011 ; Goltstein et al, 2021 ); but it remains unknown what kind of abstract knowledge (which is often represented by single or multimodal sensory stimuli) is most effective for specific species. In the case of monkey electrophysiological recordings, large unit yields may also prove crucial for the discovery of grid cells because of possibly sparse grid-cell representations.…”