Mining the Brain in an Age of Big Data

John Pearson, PhD
Duke University

Talk Abstract: Like genetics twenty years ago, neuroscience is currently anticipating the approach of a tidal wave of data. From hundreds of simultaneously recorded neurons to rich and complex social behavior, a new generation of experiments has already begun to challenge our conventional methods for visualizing, analyzing, and interpreting information. Machine learning methods have tremendous potential to address these issues, but their greatest successes have come in engineering rather than scientific applications. I will discuss several key considerations involved in applying these methods to neuroscience and illustrate with case studies including data from both humans and animal models.

Motivation as Neural Context for Learning (CANCELLED)

R. Alison Adcock, MD, PhD
Duke University

Talk Abstract: Although researchers often discuss how the brain produces behavior, it is also true that behavior and experience influence the brain. Our research has shown that distinct motivational states elicited by expectation of reward or punishment and influence learning and memory via distinct brain systems. These different motivational states correspond to differential activity and connectivity in brain circuits implicated not only in motivation but also in learning and memory. This selectivity in memory mechanisms, in turn, determines whether the information in memory is detailed versus general or flexible versus rigid. Our recent work has shown that people can self-induce activation of in neuromodulatory systems capable of broadly influencing brain function, implying that we can actively leverage these neural contexts to shape learning during education and therapy, a process we refer to as behavioral neurostimulation.

Control of patience and serotonin

Kenji Doya, PhD
Okinawa Institute of Science and Technology

Talk Abstract: Evaluation of delayed reward is an essential component of intelligence and also critical in economic decision making. We hypothesized that the brain’s serotonergic system regulates the temporal discounting of future rewards. Human brain imaging experiments showed that different cortico-basalganglia loops are involved in reward prediction in different time scales and that they are differentially modulated by  serotonerin. Chemical recording and manipulation in rats showed that serotonin release is elevated when animals are engaged in delayed reward tasks and that suppression of dorsal raphe serotonin neurons increase reward waiting errors. Electrode recording of dorsal raphe neurons revealed that their firing increases during reward waiting period and drops just before the animal abandons waiting. Optogenetic activation of dorsal raphe serotonin neurons reduced waiting errors and extended waiting time in reward omission trials. These results showed that dorsal raphe serotonin neurons control the patience for delayed rewards in a dynamic way.

Mapping spatial priority in the human frontoparietal cortex

Clay Curtis, PhD
New York University

Talk Abstract: The prefrontal and posterior parietal cortices (PFC/PPC) sit at the apex of the sensorimotor hierarchy and are important for the selection and planning of voluntary action and are thought to bias the processing in sensory areas towards behaviorally relevant dimensions. Recently, several lines of evidence from a variety of disciplines have converged on a theory positing that activity in the frontal and parietal cortices constitutes maps of prioritized space. In this conceptual framework, priority maps tag important locations in the visual field and are constructed both from the salience (e.g., conspicuousness) of objects and its current relevance (e.g., task goal). Activity in a priority map could be used to select between competing representations of actions in the motor system or between competing representations of objects in the visual system. I will describe recent efforts in my lab to test whether patterns of neural activity in the human PFC and PPC are consistent with predictions from the priority map theory. Using novel topographic mapping techniques to identify candidate priority maps in PFC and PPC, we then perform a number of experiments to test hypotheses about the nature of what is being prioritized. In general, we find that sculpted activity in topographically organized maps of prioritized space in PFC and PPC could be read out to guide a variety of spatial cognitive behaviors.

Incorporating Neurobiological Constraints Can Rationalize Human Choice Behavior

Paul Glimcher, PhD
New York University

Talk Abstract: Over the course of the last decade neuroeconomists have gathered a tremendous amount of information about the structure of neural representations. We now know how quantities are represented in the brain and what the relative costs of different forms of representation are. Neoclassical economists have spent nearly a century building theories of choice that are fundamentally theories about representation, but these utility-based theories have turned out to be incompletely predictive of human behavior. Our most recent studies indicate that one key failure of neoclassical economic theory was its failure to incorporate the costs of computation and representation into its core theoretical framework. Our theoretical and experimental work suggests that once these costs are specified (in a minimally restrictive but quite complete way) much of the hoped-for predictive power of neoclassical economics can be recovered.