Suzuki Mototaka, Pennartz Cyriel M A, Aru Jaan
Department of Cognitive and Systems Neuroscience, Swammerdam Institute for Life Sciences, University of Amsterdam, Amsterdam, The Netherlands.
Institute of Computer Science, University of Tartu, Tartu, Estonia.
Nat Rev Neurosci. 2023 Dec;24(12):778-791. doi: 10.1038/s41583-023-00756-z. Epub 2023 Oct 27.
Deep learning and predictive coding architectures commonly assume that inference in neural networks is hierarchical. However, largely neglected in deep learning and predictive coding architectures is the neurobiological evidence that all hierarchical cortical areas, higher or lower, project to and receive signals directly from subcortical areas. Given these neuroanatomical facts, today's dominance of cortico-centric, hierarchical architectures in deep learning and predictive coding networks is highly questionable; such architectures are likely to be missing essential computational principles the brain uses. In this Perspective, we present the shallow brain hypothesis: hierarchical cortical processing is integrated with a massively parallel process to which subcortical areas substantially contribute. This shallow architecture exploits the computational capacity of cortical microcircuits and thalamo-cortical loops that are not included in typical hierarchical deep learning and predictive coding networks. We argue that the shallow brain architecture provides several critical benefits over deep hierarchical structures and a more complete depiction of how mammalian brains achieve fast and flexible computational capabilities.
深度学习和预测编码架构通常假定神经网络中的推理是分层的。然而，在深度学习和预测编码架构中很大程度上被忽视的是神经生物学证据，即所有分层的皮层区域，无论高低，都直接向皮层下区域投射并接收信号。鉴于这些神经解剖学事实，深度学习和预测编码网络中以皮层为中心的分层架构如今的主导地位非常值得怀疑；这样的架构可能缺少大脑所使用的基本计算原理。在这篇观点文章中，我们提出了浅脑假说：分层的皮层处理与一个大规模并行过程相结合，皮层下区域对该过程有实质性贡献。这种浅层架构利用了典型分层深度学习和预测编码网络中未包含的皮层微回路和丘脑 - 皮层环路的计算能力。我们认为，与深度分层结构相比，浅脑架构具有几个关键优势，并且能更完整地描述哺乳动物大脑如何实现快速且灵活的计算能力。
