Researchers at University of California Berkeley Release New Data on Neural Computation (Autonomous Learning With High-Dimensional Computing Architecture Similar to Von Neumann's).

This article reports on research from the University of California Berkeley presenting a new computational model of human and animal learning using high-dimensional vectors, with dimensions around 10,000. The model’s architecture resembles traditional von Neumann computing but operates on vectors in superposition, incorporating a high-capacity memory analogous to random-access memory (RAM) and reflecting psychological and biological principles of short-term working memory and long-term storage in the cerebellar cortex. The research aims to develop a mathematical theory of vector-based computing that aligns with neuroscience and psychology, with potential applications in robotics and language processing, while emphasizing efficiency comparable to the brain’s material and energy use. The study has been peer-reviewed and published in the journal Neural Computation. [Extracted from the article]