COMPUTATIONAL MODELS OF RETINAL FUNCTION

我们研究计划的长期目标是了解神经 视觉行为的基础。我们建议通过以下方式努力实现这一目标 多孔河蟹视觉系统的研究。 这种动物的视网膜含有最大的神经网络， 存在一个量化模型。此外，我们理解它的视觉效果 行为足够好，可以精确地确定它能看到什么。我们 计划将理论和行为相结合来研究神经机制 控制视觉行为的处理。我们希望了解到什么 眼睛必须发送给大脑的信息才能让动物看到。 具体目标是调查： 。大规模并行的视网膜功能计算模型 电脑 。行为背后的视觉信息的神经编码 我们计划进行一项系统的研究，包括计算、行为和 生理学。到目前为止，我们已经编制了一个实际的 在大规模并行计算机上的1000神经元视网膜，连接 机器(CM5型)。为了验证这个模型，我们开发了一种技术 用于记录表现良好的动物的单个视网膜神经元。这些 这些技术非常适合研究信息处理。 并将神经编码与视觉表现联系起来。他们 为本研究建立一个合适的动物模型。 过去对河豚的研究为视觉研究提供了里程碑。 拟议的研究应在以下领域证明同样富有成效 神经编码和计算神经科学。他们承诺会揭晓 大规模神经网络中信息处理的基本机制 从而有助于我们理解神经元是如何控制行为的。

The long-range goal of our research program is to understand the neural basis of visual behavior. We propose to work toward this goal by studying the visual system of the horseshoe crab, Limulus Polyphemus. The retina of this animal contains the largest neural network for which a quantitative model exists. In addition, we understand its visual behavior well enough to determine with precision what it can see. We plan to combine theory and behavior to study the mechanisms of neural processing that control visual behavior. We hope to learn what information the eye must send to the brain for the animal to see. Specific aims are to investigate: . Computational models of retinal function on a massively parallel computer . Neural coding of visual information underlying behavior We plan a systematic study involving computation, behavior, and physiology. To date we have programmed a preliminary model of the actual 1000-neuron retina on a massively parallel computer, the Connection Machine (Model CM5). To verify the model we have developed a technique for recording from single retinal neurons of a behaving animal. These techniques are ideally suited for investigating information processing in the retina and relating neural coding to visual performance. They establish Limulus as an appropriate animal model for this research. Past studies with Limulus have provided milestones in vision research. The proposed studies should prove equally productive in the areas of neural coding and computational neuroscience. they promise to reveal basic mechanisms of information processing in a sizable neural network and thus contribute to our understanding of how neurons control behavior.