Brain Computation as Hierarchical Abstraction (Computational Neuroscience Series)

The vast differences between the brain's neural circuitry and a computer's silicon circuitry might suggest that they have nothing in common. In fact, as Dana Ballard argues in this book, computational tools are essential for understanding brain function. Ballard shows that the hierarchical organization of the brain has many parallels with the hierarchical organization of computing; as in silicon computing, the complexities of brain computation can be dramatically simplified when its computation is factored into different levels of abstraction.

Drawing on several decades of progress in computational neuroscience, together with recent results in Bayesian and reinforcement learning methodologies, Ballard factors the brain's principal computational issues in terms of their natural place in an overall hierarchy. Each of these factors leads to a fresh perspective. A neural level focuses on the basic forebrain functions and shows how processing demands dictate the extensive use of timing-based circuitry and an overall organization of tabular memories. An embodiment level organization works in reverse, making extensive use of multiplexing and on-demand processing to achieve fast parallel computation. An awareness level focuses on the brain's representations of emotion, attention and consciousness, showing that they can operate with great economy in the context of the neural and embodiment substrates.

大腦的神經電路和電腦的矽電路之間的巨大差異可能表明它們沒有共同之處。事實上，正如 Dana Ballard 在這本書中所主張的，計算工具對於理解大腦功能至關重要。Ballard 表明，大腦的階層組織與計算的階層組織有許多相似之處；就像矽計算一樣，當大腦的計算被分解為不同的抽象層次時，大腦計算的複雜性可以被大大簡化。

Ballard 借鑒了數十年來在計算神經科學中的進展，以及貝葉斯和強化學習方法論的最新成果，將大腦的主要計算問題分解為其在整體層次結構中的自然位置。這些因素中的每一個都帶來了新的視角。神經層次關注基本的前腦功能，並展示了處理需求如何決定大量使用基於時間的電路和整體記憶的組織。體現層次組織則相反，大量使用多路復用和按需處理來實現快速並行計算。意識層次關注大腦對情緒、注意力和意識的表徵，並顯示它們在神經和體現基板的背景下可以以極高的經濟性運作。