Pei X, Moss F
Department of Physics and Astronomy, University of Missouri, St. Louis 63121, USA.
Int J Neural Syst. 1996 Sep;7(4):429-35. doi: 10.1142/s0129065796000403.
We discuss the well-known problems associated with efforts to detect and characterize chaos and other low dimensional dynamics in biological settings. We propose a new method which shows promise for addressing these problems, and we demonstrate its effectiveness in an experiment with the crayfish sensory system. Recordings of action potentials in this system are the data. We begin with a pair of assumptions; that the times of firings of neural action potentials are largely determined by high dimensional random processes or "noise"; and that most biological files are non stationary, so that only relatively short files can be obtained under approximately constant conditions. The method is thus statistical in nature. It is designed to recognize individual "events" in the form of particular sequences of time intervals between action potentials which are the signatures of certain well defined dynamical behaviors. We show that chaos can be distinguished from limit cycles, even when the dynamics is heavily contaminated with noise. Extracellular recordings from the crayfish caudal photoreceptor, obtained while hydrodynamically stimulating the array of hair receptors on the tailfan, are used to illustrate the method.
我们讨论了在生物环境中检测和表征混沌及其他低维动力学所涉及的一些众所周知的问题。我们提出了一种有望解决这些问题的新方法，并在小龙虾感官系统的实验中证明了其有效性。该系统中动作电位的记录即为数据。我们首先提出两个假设：神经动作电位的发放时间很大程度上由高维随机过程或“噪声”决定；并且大多数生物记录是非平稳的，因此在近似恒定的条件下只能获得相对较短的记录。所以该方法本质上是统计性的。它旨在识别以动作电位之间特定时间间隔序列形式存在的个体“事件”，这些时间间隔序列是某些明确定义的动力学行为的特征。我们表明，即使动力学被大量噪声污染，也能将混沌与极限环区分开来。通过对小龙虾尾端光感受器进行细胞外记录（在对尾扇上的毛感受器阵列进行流体动力学刺激时获得）来说明该方法。
