Cerebellar strategies of supporting diverse functions: a tripartite computational framework

Background Despite its uniform cytoarchitecture, the cerebellum coordinates both sensorimotor adaptation and higher-order cognitive functions. How a canonical circuit executes such divergent operations is debated between the Universal Cerebellar Transform hypothesis (an invariant predictive algorithm across domains) and the Task-Specific Transformation hypothesis (hidden cellular, molecular, and electrophysiological heterogeneities tuning local microcircuits).Methods We propose a conceptual tripartite framework integrating three complementary learning principles in cerebellar computation: (1) unsupervised learning in the granular layer via high-dimensional sparse expansion; (2) supervised learning via climbing fiber-instructed synaptic plasticity minimizing sensory prediction errors; and (3) reinforcement learning mediated by reward-encoding climbing fibers and cerebello-mesolimbic loops.Results and Conclusion We map these strategies onto specific circuit components, examine internal forward and inverse models across motor and cognitive domains, and discuss open questions regarding climbing fiber multiplexing, memory consolidation, and bio-inspired neuromorphic engineering.