Flexible brain-inspired hybrid analog-spiking neuronal network computation in energy-efficient neuromorphic hardware (FlexBrain)
FlexBrain is one of the projects supporting advancements in artificial intelligence for science funded by the U.S. Department of Energy (DOE).
Brains exhibit discrete spiking and analog dynamics. The latter, including neuromodulation and graded communication, shape collective brain state dynamics attributed to computational coordination. This project hypothesizes that the brain’s ability to flexibly adapt computations “on the fly” and perform energy-efficient intelligent computation emerges from a synergy between discrete spiking and analog dynamics.
The goal of FlexBrain project is to develop a brain-inspired neuromorphic computing framework, which employs a hybrid combination of spiking neural networks (SNN) for energy-efficient large-scale computations and analog neural oscillator networks that mimic brain state-dependent computational coordination dynamics.
CXRO’s Role
The CXRO team works on applying the developed hybrid artificial neural networks for the online analysis of large image data streams, including X-ray image processing and the application of machine learning algorithms for nanomaterials research. CXRO provides training data generated from experiments and micromagnetic simulations.
Partner Institutions
- Lawrence Berkeley National Laboratory (CRD, ALS)
- University of San Francisco
- University of Southern California
