Artificial neural manifolds.

Artificial neural manifolds based on Mott memristors enable rapid, accurate prediction with fewer samples, mimicking biological neural structures.

• Why it matters: Understanding and replicating the brain’s ability to perform perception and prediction is crucial for advancing neuromorphic systems, yet most current models ignore the underlying manifold structure.
• What they did: Researchers built an artificial neuron circuit that produces biologically similar bell-shaped tuning curves, forming a low-dimensional manifold that supports spatiotemporal prediction using a modified STI equation with a memory factor.
• The result: This approach achieves precise perception of incomplete images and accurate prediction of epileptic seizures, enhancing robustness and efficiency in neural information processing systems.