Top-notch researchers at the University of Surrey have unveiled a brain-inspired technique called Topographical Sparse Mapping to wire neural networks more efficiently and boost AI performance.

The study, published in Neurocomputing, provides peer-reviewed validation of this approach.

An enhanced variant, Enhanced Topographical Sparse Mapping, introduces a biologically inspired pruning process during training to further refine neural connections.

Looking ahead, researchers plan to explore broader applications and implications of brain-inspired topology for AI and computing architectures.

The team emphasizes that current large AI models demand substantial electricity, and their method could lower energy needs without sacrificing performance.

The approach improves generative AI and other modern models while preserving accuracy and cutting energy consumption.

Beyond software, the team is exploring neuromorphic computing applications that aim to build hardware mirroring brain structure and function.

The core idea is to connect each neuron only to nearby or related neurons, mirroring the brain’s efficient, sparse information wiring to reduce unnecessary connections and energy use.