Research Projects

Project Summary The field of artificial intelligence (AI) has recently made remarkable advances that resulted in new and improved algorithms and network architectures that proved efficient empirically in silico. These advances raise new questions in neurobiology: are these new algorithms used in the brain? The present study focuses on a new algorithm developed in the field of reinforcement learning (RL), called distributional RL, which outperforms other state-of-the-art RL algorithms and is regarded as a major advancement in RL.

ABSTRACT Our long-term goal is to generate a complete understanding of how the cerebellum learns to improve movement in response to motor errors. Climbing fibers are thought to play an essential role in this process because they fire during erroneous movement. Their activity reliably excites Purkinje cells, eliciting calcium spikes in their dendrites that can trigger long-term synaptic plasticity at coactive parallel fiber inputs. Plasticity induction ultimately leads to corrective behavior by altering the cerebellum’s response to sensorimotor stimuli that predict mistakes.

Project Abstract Understanding how neuronal computations build up a perception of the external world is fundamental to our understanding of how the brain works. This is particularly relevant to sensory systems, where heterogenous inputs representing distinct sensory features must be re-assembled to generate a perception. How individual neurons in early stages of sensory circuits process parallel inputs, and how these circuit elements later contribute to cortical computations that bind the inputs together is completely unknown.

7. Project Summary/Abstract Human experience is shaped by our senses, which receive diverse inputs from our environment. These varied inputs, however, all contribute to a single integrated representation in our minds of the outside world. Neuroscientists have studied sensory perception, and the integration different sensory modalities like vision, hearing, and touch, for decades, but there are still important unanswered questions about how sensory cells in the brain work, and how the circuits that they form control the flow of information through the brain.

Work on learning in neural systems has focused largely on the effects of plasticity at synapses that provide direct input to the neurons being studied. Learning a model of the environment or a complex skill, however, relies on plasticity that is widely distributed and may occur at synapses far from the neurons driving decisions or actions. As is well-known from multi-layer (or 'deep') artificial networks, distributing learning over multiple layers is substantially more powerful but also more difficult to implement than learning at a single layer.

Abstract Our goal is to generate a predictive model of the spinal cord nociceptive circuits that underlie the initiation of pain perception and behavior. Nociceptive signals are conveyed from the periphery to the spinal cord dorsal horn via highly specialized primary sensory neuron subtypes. These sensory neurons, as well as descending modulatory neurons, form synapses upon an array of morphologically and physiologically distinct classes of dorsal horn interneuron and projection neurons.

ABSTRACT Cephalopods have large and complex brains, and in particular a highly capable visual system. However, their brains evolved independently from vertebrates, and very little is known about how neural circuits in the cephalopod brain process visual information. In fact, there has never been a direct recording of receptive fields in the central visual system of cephalopods. This study will measure neural activity and visual coding in the optic lobe of Octopus bimaculoides, an emerging model organism for cephalopod research.

Sensory processing requires the interaction between external inputs and an internal brain state. Pain is a unique sensory experience that is triggered by external signals, but is also strongly shaped by internal cognitive and emotional variables. At the circuit level, there is not a single primary pain cortex; instead, a distributed network of cortical areas process and regulate pain. For example, the primary somatosensory cortex (S1) is known to process stimulus-evoked information, such as location and timing.

Project Summary [30 lines max] Timing is critical to neural processing. Nowhere is that clearer than in visual motion detection. To detect motion, neurons transmit visual information with different latencies, or delays, allowing the circuit to compare visual scenes over time. When comparisons over time are combined with comparisons over space, the circuit can compute direction-selective signals, which are larger for motion in one direction than in the opposite direction. These signals in turn guide a wide range of behaviors, from navigation and predator avoidance to mating.

Project Summary Here, we propose to thoroughly characterize the origins of pairwise correlations in cortex using a synergistic mix of experimental methodologies, behavior, and computation in mice and macaques. We will elucidate the mechanistic underpinnings of normalization and test our hypothesis that changes in cortical pairwise correlations and other signature arise from ongoing cortical computations.