Effects of Noise on the Electrosensory System of Mormyrid Electric Fish

大脑处理连续的感官信息流，但生理信号中也存在噪声，影响信息处理的不确定性。突触是神经细胞（神经元）之间的功能性接触，在突触的微观层面上，相关信号是称为突触电位的小变量电生理事件。当学习发生时，这些突触电位的相对加权值会发生变化，即它们对目标神经元的影响有多大，从而激发并最终驱动特定行为。在大脑功能区域更宏观的层面上，存在对大量细胞输入作出反应的系统特性。该系统需要在准确性和适应性之间进行优化权衡，这可能需要一些时间来建立，而适应性通常需要快速调整以适应感官环境的变化。该合作项目将神经网络方法与计算生物学和实验神经生理学相结合，以测试自适应学习规则的统计模型。实验涉及一种弱电鱼，它在自身周围产生脉冲电场。水中的物体会扭曲磁场，身体上的侧线传感器会检测到这些扭曲。他们大脑中电感觉侧线叶（ELL）的神经元接收来自侧线传感器的信号，同时也接收暂时锁定在电脉冲运动指令上的信号。ELL细胞的反应适应不断变化的电感觉条件，这种适应性使这些神经元能够“存储”鱼对电感觉场的期望图像。目前还不清楚在细胞水平上看到的适应性学习是否可以解释行为鱼类大脑中的集体神经活动。这个项目涉及一个新的建模方面，强调在噪声存在下突触系统的统计特性，并对噪声如何影响鱼对信号的存储期望的准确性做出可测试的预测。该项目的结果将阐明突触学习率和信号处理如何相互作用，以及系统如何处理噪声的影响。其影响将超越电感觉处理和计算神经科学，扩展到学习和小脑样大脑结构如何运作的研究，并涉及高度跨学科的培训。

The brain handles a continuous temporal flow of sensory information, but there also is noise in the physiological signal that affects the uncertainty in information processing. At the microscopic level of the synapses, which are the functional contacts between nerve cells (neurons), the relevant signals are the small variable electrophysiological events called synaptic potentials. When learning occurs, there is a change in the relative weighted value of these synaptic potentials, for how strongly they influence the targeted neurons to fire and ultimately to drive a particular behavior. At the more macroscopic level of a functional region of the brain, there are systems properties of the response to large numbers of cellular inputs. The system needs to optimize the trade-off between accuracy, which may take some time to establish, and adaptability, which often requires rapid adjustment to changes in the sensory environment. This collaborative project combines a neural network approach with computational biology and experimental neurophysiology, to test a statistical model on rules of adaptive learning. Experiments involve a species of weakly electric fish, which produces a pulsed electrical field around itself. Objects in the water distort the field, and the lateral-line sensors on the body detect these distortions. Neurons in the electrosensory lateral-line lobe (ELL) in their brain receive signals from the lateral line sensors, and also receive signals temporally locked to the motor command for the electric pulses. Responses of cells in the ELL adapt to changing electrosensory conditions, and this adaptability allows these neurons to 'store' an image of the fish's expectation of the electrosensory field. It is not clear whether the adaptive learning seen at the cellular level can explain collective neural activity in the brain of the behaving fish. This project involves a novel aspect of modeling emphasizing statistical properties of the synaptic system in the presence of noise, and making testable predictions about how noise can affect the accuracy of the fish's stored expectation of the signal. Results of this project will clarify how synaptic learning rates and signal processing interact, and how the effects of noise are handled by the system. The impact will extend beyond electrosensory processing and computational neuroscience to studies of learning and of how cerebellar-like brain structures function, and there is highly cross-disciplinary training involved.