Mathematical Sciences:Mathematical Modeling of Neural Populations

研究人员Abbott研究了大量神经元的动态特性如何允许它们表示、存储、回忆和处理信息。该项目的一个基本主题是研究群体编码，即在大量神经元阵列中存储信息，每个神经元对刺激的不同方面做出相当非选择性的反应。用种群译码的方法研究了神经元网络之间的信息传递。要解决的问题包括：在视觉引导向物体伸展之类的任务中，感觉网络中编码的信息如何传输到运动网络？独立于平移、旋转和缩放变化来识别和定位对象所需的不变量表示是如何产生的？人们普遍认为，突触修饰是记忆和学习的基本神经机制。群体解码方法允许用直接的行为相关性来解释突触改变的影响。在初步的解码研究中，人们发现，长时程增强的时间特性自然会导致一群神经元在训练后预测编码量。这表明了一种新的学习和生成运动动作序列的机制被探索。伸手拿物体，转向视觉或听觉提示，以及各种其他任务，都需要由感觉信息准确引导的运动反应。此外，以前的经验可以在形成这些反应方面发挥重要作用。实验提供了有关感觉信息如何在大脑中呈现以及经验如何改变神经回路的重要信息。数学解码技术使我们能够将伸展等运动反应与大脑运动皮质神经元的活动联系起来。该项目的目标是在这些数据的基础上构建一个理论框架，用于理解感官引导的运动任务，包括学习和产生准确反应并根据经验修改它们的机制。这项工作应该会为我们如何产生基本的运动行为提供新的见解，并可能在机器人学中有有趣的应用。

Abbott The investigator studies how the dynamic properties of large populations of neurons allow them to represent, store, recall and process information. A basic theme of the project is the study of population coding, the storage of information in large arrays of neurons, each responding fairly nonselectively to different aspects of a stimulus. The transfer of information between neuronal networks is studied using population decoding methods. Questions to be addressed include: How is the information encoded in a sensory network transferred to a motor network in a task like visually guided reaching toward an object? How do the invariant representations needed to identify and locate an object independent of translations, rotations, and scaling changes arise? It is widely believed that synaptic modification is the basic neuronal mechanism underlying memory and learning. Population decoding methods allow for an interpretation of the effects of synaptic modification with direct behavioral relevance. In preliminary decoding studies it has been found that the temporal properties of long-term potentiation naturally cause a population of neurons to predict a coded quantity after training. This suggests a novel mechanism for learning and generating sequences of motor actions that is explored. Reaching for an object, turning toward a visual or auditory cue, and a variety of other tasks require motor responses that are accurately guided by sensory information. In addition, previous experience can play an important role in shaping these responses. Experiments have provided important information about how sensory information is represented in the brain and how experience can modify neural circuits. Mathematical decoding techniques allow us to relate motor responses, such as reaching, to the activity of neurons in the motor cortex of the brain. The goal of the project is to construct a theoretical framework, based on these data, for understanding sensory-guid ed motor tasks, including mechanisms for learning and generating accurate responses and modifying them on the basis of experience. This work should provide new insights into how we generate basic motor behaviors and may have interesting applications to robotics.