Locomotor action sequences impact the scale of representation in hippocampus and posterior parietal cortex

Local prediction-learning in high-dimensional spaces enables neural networks to plan

Planning and problem solving are cornerstones of higher brain function. But we do not know how the brain does that. We show that learning of a suitable cognitive map of the problem space suffices. Furthermore, this can be reduced to learning to predict the next observation through local synaptic plasticity. Importantly, the resulting cognitive map encodes relations between actions and observations, and its emergent high-dimensional geometry provides a sense of direction for reaching distant goals. This quasi-Euclidean sense of direction provides a simple heuristic for online planning that works almost as well as the best offline planning algorithms from AI. If the problem space is a physical space, this method automatically extracts structural regularities from the sequence of observations that it receives so that it can generalize to unseen parts. This speeds up learning of navigation in 2D mazes and the locomotion with complex actuator systems, such as legged bodies. The cognitive map learner that we propose does not require a teacher, similar to self-attention networks (Transformers). But in contrast to Transformers, it does not require backpropagation of errors or very large datasets for learning. Hence it provides a blue-print for future energy-efficient neuromorphic hardware that acquires advanced cognitive capabilities through autonomous on-chip learning.

Wide-Field Calcium Imaging of Dynamic Cortical Networks during Locomotion

Motor behavior results in complex exchanges of motor and sensory information across cortical regions. Therefore, fully understanding the cerebral cortex's role in motor behavior requires a mesoscopic-level description of the cortical regions engaged, their functional interactions, and how these functional interactions change with behavioral state. Mesoscopic Ca2+ imaging through transparent polymer skulls in mice reveals elevated activation of the dorsal cerebral cortex during locomotion. Using the correlations between the time series of Ca2+ fluorescence from 28 regions (nodes) obtained using spatial independent component analysis (sICA), we examined the changes in functional connectivity of the cortex from rest to locomotion with a goal of understanding the changes to the cortical functional state that facilitate locomotion. Both the transitions from rest to locomotion and from locomotion to rest show marked increases in correlation among most nodes. However, once a steady state of continued locomotion is reached, many nodes, including primary motor and somatosensory nodes, show decreases in correlations, while retrosplenial and the most anterior nodes of the secondary motor cortex show increases. These results highlight the changes in functional connectivity in the cerebral cortex, representing a series of changes in the cortical state from rest to locomotion and on return to rest.