Institute of Cognitive Sciences and Technologies, National Research Council, Via San Martino della Battaglia 44, 00185 Rome, Italy.
Department of Psychology, University of California at San Diego, La Jolla, CA 92093, USA.
Neurosci Biobehav Rev. 2019 May;100:19-34. doi: 10.1016/j.neubiorev.2019.02.008. Epub 2019 Feb 18.
Despite wide evidence suggesting anatomical and functional interactions between cortex, cerebellum and basal ganglia, the learning processes operating within them --often viewed as respectively unsupervised, supervised and reinforcement learning-- are studied in isolation, neglecting their strong interdependence. We discuss how those brain areas form a highly integrated system combining different learning mechanisms into an effective super-learning process supporting the acquisition of flexible motor behaviour. The term "super-learning" does not indicate a new learning paradigm. Rather, it refers to the fact that different learning mechanisms act in synergy as they: (a) affect neural structures often relying on the widespread action of neuromodulators; (b) act within various stages of cortical/subcortical pathways that are organised in pipeline to support multiple sensation-to-action mappings operating at different levels of abstraction; (c) interact through the reciprocal influence of the output compartments of different brain structures, most notably in the cerebello-cortical and basal ganglia-cortical loops. Here we articulate this new hypothesis and discuss empirical evidence supporting it by specifically referring to motor adaptation and sequence learning.
尽管有广泛的证据表明大脑皮层、小脑和基底神经节之间存在解剖和功能上的相互作用,但在这些区域内运作的学习过程——通常被视为分别为无监督学习、监督学习和强化学习——却被孤立地研究,而忽略了它们之间的强烈相互依赖关系。我们讨论了这些脑区如何形成一个高度整合的系统,将不同的学习机制结合成一个有效的超级学习过程,以支持灵活的运动行为的习得。“超级学习”一词并不表示一种新的学习范式。相反,它指的是不同的学习机制协同作用的事实,因为它们:(a) 影响神经结构,通常依赖于神经调质的广泛作用;(b) 在皮质/皮质下通路的不同阶段发挥作用,这些通路被组织成流水线,以支持在不同抽象层次上运作的多个感觉-运动映射;(c) 通过不同脑结构的输出隔室的相互影响进行交互,特别是在小脑-皮质和基底神经节-皮质环路中。在这里,我们阐述了这一新假设,并通过具体参考运动适应和序列学习来讨论支持它的经验证据。