Explore the vast range of titles available.

Reinforcement has long been thought to require striatal synaptic plasticity. Indeed, direct striatal manipulations such as self-stimulation of direct-pathway projection neurons (dMSNs) are sufficient to induce reinforcement within minutes. However, it’s unclear what role, if any, is played by downstream circuitry. Here, we used dMSN self-stimulation in mice as a model for striatum-driven reinforcement and mapped the underlying circuitry across multiple basal ganglia nuclei and output targets. We found that mimicking the effects of dMSN activation on downstream circuitry, through optogenetic suppression of basal ganglia output nucleus substantia nigra reticulata (SNr) or activation of SNr targets in the brainstem or thalamus, was also sufficient to drive rapid reinforcement. Remarkably, silencing motor thalamus—but not other selected targets of SNr—was the only manipulation that reduced dMSN-driven reinforcement. Together, these results point to an unexpected role for basal ganglia output to motor thalamus in striatum-driven reinforcement.

Complex motor skills are more difficult to perform at certain points in the day (for example, shortly after waking), but the daily trajectory of motor-skill error is more difficult to predict. By undertaking a quantitative analysis of the fundamental frequency (FF) and amplitude of hundreds of zebra finch syllables per animal per day, we find that zebra finch song follows a previously undescribed daily oscillation. The FF and amplitude of harmonic syllables rises across the morning, reaching a peak near mid-day, and then falls again in the late afternoon until sleep. This oscillation, although somewhat variable, is consistent across days and across animals and does not require serotonin, as animals with serotonergic lesions maintained daily oscillations. We hypothesize that this oscillation is driven by underlying physiological factors which could be shared with other taxa. Song production in zebra finches is a model system for studying complex learned behavior because of the ease of gathering comprehensive behavioral data and the tractability of the underlying neural circuitry. The daily oscillation that we describe promises to reveal new insights into how time of day affects the ability to accomplish a variety of complex learned motor skills.

The control of locomotion is fundamental to vertebrate animal survival. Defensive situations require an animal to rapidly decide whether to run away or suppress locomotor activity to avoid detection. While much of the neural circuitry involved in defensive action selection has been elucidated, top-down modulation of brainstem locomotor circuitry remains unclear. Here we provide evidence for the existence and functionality of a monosynaptic connection from the central amygdala (CeA) to the mesencephalic locomotor region (MLR) that inhibits locomotion in unconditioned and conditioned defensive behavior in mice. We show that locomotion stimulated by airpuff coincides with increased activity of MLR glutamatergic neurons. Using retrograde tracing and ex vivo electrophysiology, we find that the CeA makes a monosynaptic connection with the MLR. In the open field, in vivo stimulation of this projection suppressed spontaneous locomotion, whereas inhibition of this projection had no effect. However, inhibiting CeA terminals within the MLR increased both neural activity and locomotor responses to airpuff. Finally, using a conditioned avoidance paradigm known to activate CeA neurons, we find that inhibition of the CeA projection increased successful escape, whereas activating the projection reduced escape. Together these results provide evidence for a new circuit substrate influencing locomotion and defensive behaviors.

We explore using neural operators, or neural network representations of nonlinear maps between function spaces, to accelerate infinite-dimensional Bayesian inverse problems (BIPs) with models governed by nonlinear parametric partial differential equations (PDEs). Neural operators have gained significant attention in recent years for their ability to approximate the parameter-to-solution maps defined by PDEs using as training data solutions of PDEs at a limited number of parameter samples. The computational cost of BIPs can be drastically reduced if the large number of PDE solves required for posterior characterization are replaced with evaluations of trained neural operators. However, reducing error in the resulting BIP solutions via reducing the approximation error of the neural operators in training can be challenging and unreliable. We provide an a priori error bound result that implies certain BIPs can be ill-conditioned to the approximation error of neural operators, thus leading to inaccessible accuracy requirements in training. To reliably deploy neural operators in BIPs, we consider a strategy for enhancing the performance of neural operators, which is to correct the prediction of a trained neural operator by solving a linear variational problem based on the PDE residual. We show that a trained neural operator with error correction can achieve a quadratic reduction of its approximation error, all while retaining substantial computational speedups of posterior sampling when models are governed by highly nonlinear PDEs. The strategy is applied to two numerical examples of BIPs based on a nonlinear reaction--diffusion problem and deformation of hyperelastic materials. We demonstrate that posterior representations of the two BIPs produced using trained neural operators are greatly and consistently enhanced by error correction.