High-Dimensional Representations in Brain and Machine: Vector Symbolic Architectures for Neural Computation

Recent advances in neuroscience, cognitive science, and artificial intelligence are converging on the need for representations that are at once distributed, interpretable, and compositional. Vector Symbolic Architectures (VSA), also known as Hyperdimensional Computing (HDC), provide a unifying framework that addresses this need by encoding structured information in high-dimensional vectors through well-defined algebraic operations. This special issue brings together contributions from across disciplines to examine VSAs as a bridge between neural computation, cognitive models of memory and reasoning, and the design of efficient, generalizable AI systems. Topics include the role of VSAs in modeling variable binding and associative memory in the brain, connections to attention and memory mechanisms in large artificial neural networks, theoretical links to sketching and random projection methods, and emerging hardware implementations in neuromorphic and in-memory computing. By integrating perspectives from cognition, neuroscience, and machine learning, this issue aims to catalyze new research directions for training-free compositionality, memory-augmented neural architectures, and energy-efficient computation.