NC4 will focus on the algorithmic role of neural circuits in controlling animal behavior. In particular, we will quantify the computational transformations effected by neural networks in higher-level cognitive regions of the brain and how these regions are integral to the closed-loop control of behavior. We will examine these circuits through an engineer’s lens, using the language of control theory and robotics to describe their computations, while respecting the vast difference in mechanisms between engineered and biological systems.

In particular, the initial focus of the lab will be spatial navigation. Robots and animals including humans can navigate their environment and perform tasks. Robotic algorithms frequently utilize a map-like representation of the world to solve and execute navigational tasks. In mammals, the hippocampal formation is the locus of a similar ‘cognitive map’. The high-level questions are:

  • What is the relationship between the structure of the map and the sensory inputs available to the animal, and how is this structure influenced by the task that the animal is being trained to perform?
  • How is the internal representation of the world, the cognitive map, being used to compute and perform navigational behaviors? Does the structure of the map facilitate this computation?

Using virtual reality, we will modulate multiple sensory inputs available to rodents, while performing electrophysiological recordings from large populations of neurons in the hippocampal formation. These experiments will be performed while the animals perform complex navigational challenges. The modulations being applied will be closed-loop, tied to the animal’s own behavioral or even neural responses. These recordings will be analyzed at the level of individual neural responses, local neural circuits, and larger circuits that span different brain regions. The goal is two-fold: (1) understand the relationship between the behavioral task and the topology of the cognitive representations, and (2) describe the transformation between cognitive representations (the neural ‘input’) and navigation (the behavioral ‘output’).

The brain is a feedback control system, with a plethora of hidden states both persistent and plastic. Higher-order representations in the brain are influenced by sensory inputs and task perception, but are driven primarily through persistent internal dynamics. It is the gradual and continuous perturbation of these neural dynamics through sensory modification which could yield the answer to their structure and function.

Quantifying the algorithmic basis of spatial navigation is a challenge that has implications well beyond understanding how a rat can find its food in a maze. Cognitive representations are integral to learning and memory, and the brain regions housing them are affected through neurodegenerative dementias such as Alzheimer’s and Huntington’s diseases. Understanding spatial navigation in rodents is key to understanding problem solving and memory in humans, and likely illuminates a path to early detection and better management of chronic degenerative disorders.