CRI: Reliable and Adaptive Behavior of Neurons: Experiment and Theory

在过去的十年里，以理解大脑为目的的计算研究有了巨大的增长。这在很大程度上是由于计算机能力的巨大扩展，以及在研究类似神经的现象（如模式识别和记忆形成）方面有前途的新程序的结果。在此期间，我们对细胞和电路水平的小神经系统的知识也有了平行的增长。因为大脑在细胞数量和它们之间的联系方面是如此复杂，较小的无脊椎动物神经系统，其运作方式与较大的神经系统完全相同，可以作为理解信息处理基本特性的模型。现在有许多小的电路已经被完全描述为突触连接。虽然这样的电路可以用现象学来建模，但人们对它们运作的基本原理缺乏深刻的理解。我们的方法是在龙虾的30个细胞的口胃神经系统中寻找这样的一般原理。我们将使用在物理学中非常成功的新数学技术，对神经节产生的有节奏的运动输出模式进行动态的非线性分析。这些方法尚未应用于单个已识别神经元和小神经回路水平的神经系统研究。这个项目既特别又及时，因为它有能力在理论和实验之间来回移动，其分辨率在其他系统中是不可能的。利用这种方法，我们希望为神经网络如何产生和处理信息提供一种新的、可概括的描述，从而为动物实现可靠、稳健和自适应的行为。我们相信，我们在这里描述的跨学科协同作用，是机器人和人工智能未来发展的关键，同时也提供了有关神经处理的基本信息。

IBN: 9634405 PI: Abarbanel The last ten years has seen a huge increase in the number of computational studies aimed at understanding the brain. This is largely the result of an enormous expansion of computer power as well as promising new programs which are useful in studying neural-like phenomena such as pattern recognition and memory formation. During this period there has also been a parallel increase in our knowledge of small nervous systems at the cellular and circuit level. Because the brain is so complex in terms of the number of cells and connections between them, smaller invertebrate nervous systems, which operate in exactly the same manner as larger ones, can serve as models for understanding fundamental properties of information processing. There are now many small circuits which have been completely described in terms of synaptic connectivity. While such circuits can be modeled phenomenologically, there is a lack of deep understanding of the fundamental principles by which they operate. Our approach will be to look for such general principles in the thirty celled stomatogastric nervous system of the lobster. We will use new mathematical techniques which have been so successful in physics, to provide a dynamical, nonlinear analysis of the rhythmic motor output patterns generated by the ganglion. These methods have not yet been applied to studies of the nervous system at the level of single identified neurons and small neural circuits. This project is both special and timely because it has the ability to move back and forth between theory and experiment at a level of resolution not possible in other systems. Using this approach, we hope to provide a new and generalizable description of how neural networks produce and process information so as to achieve reliable, robust and adaptive behavior for the animal. We believe a cross-disciplinary synergy such as we describe here, holds the key for future developments in robotics and artificial intelligence as well as providin g basic information about neural processing.