Neural and Cognitive Systems

Unlike traditional approaches based on connectionism, which relies on symbolic manipulation and encoding of representations (passive approach), our focus is on large-scale modeling of the nervous system and the interaction of nervous system, body, and the environment (active approach) as a basis for the emergence of cognitive functions.

In this active approach, internal representations are emergent from actions in a roughly Piagetian sense, where emphasis is placed on timed coordination of actions and interactions with the environment. This coordination is facilitated based on a computational architecture composed of neuronal oscillators and synapses with timing-dependent adaptive mechanisms (such as spike-timing-dependent plasticity) and conduction delays as its basic functional unit.

We believe functional relationships are rendered via oscillations at these basic functional units and their modulation via neuromodulatory influences. We aim to develop a computing framework that simulates these features with a high degree of fidelity to enable learning of complex, intelligent behavior when interacting with a rich, dynamic environment.

CNES is exploring brain function in a series of programs aimed at producing detailed models of the brain at several levels of description:

  • Hardware implementations of spiking neurons that make it practical to simulate brain regions at unprecedented scale and accuracy, even to the size of a cat brain.
  • Software simulations at the level of rate-coded neurons, detailing complex dynamic interactions between executive control, memory, posterior cortices, and neuromodulation.
  • Software functional models of cognitive processes in the form of production rules and chunks of declarative knowledge.

Models of brain function must be immersed in realistic situations for learning and validation, so a critical part of CNES activity is to develop a set of applications in the area of bio-inspired robotics.

Sections

Program and Sponsor

Project: Systems of Neuromorphic Adaptive Plastic Scalable Electronics (SyNAPSE)
Sponsor: Defense Advanced Research Projects Agency (DARPA)
Partners: Neurosciences Institute, Boston University, University of California Irvine, Georgia Institute of Technology, George Mason University, University of Michigan, SET Corporation, Portland State University, University of Nevada Reno