Analyzing Neural Activity Using Information Theory

描述(改编自申请人的摘要)：如果 神经系统是信息处理的，那么我们应该能说什么 它意味着生物系统处理信息。不幸的是， 几乎不能用数量的方式来说。要解决这个问题，请尝试 在过去的50年里断断续续地被用来理解大脑 通过应用信息论(IT)来发挥作用。近年来，这样的尝试 已经变得更加频繁，偶尔也会取得有趣的成功。然而， 从IT工程学科的角度来看，这些尝试有 几乎没有触及IT所提供的东西的皮毛，到目前为止，人们已经 仅尝试将最广为人知和最基本的方面 它。 事实上，应用了明显成功的思想--源代码编码理论--是 基本上和香农的原著一样古老。从那时起，许多人 发生在IT领域，包括香农工作的几个扩展。在这里， 调查人员主张引入RD理论及其最新成果， 逐次求精理论。拟议研究的目标是 创建一个演示，说明其适用性和前景 这些更复杂的IT成果的优越性。调查人员将 翻译率失真理论和逐次求精理论的思想 从它们产生的传播文献，到 生物计算。这种翻译必然是抽象的， 数学上的。但与此同时，调查人员将进一步展示 人们如何应用这些见解以及检验几个猜想。 出于生物动机的小例子的计算机模拟，例子很好 在神经网络理论的范围内，以及 研究认知的计算基础，将被用来说明 理论正在发展中。最后，研究人员描述了RD理论和 逐次精炼的翻译版本中固有的动态 理论可以用来量化一些最普遍的神经隐喻。 计算。因此，尽管使用已知的 IT的结果，研究人员认为神经科学家必须从 应用该理论的一些更深层次的方面。尤其是，过度的 现在神经科学文献中存在的对IT的简单化使用必须 明晰升级。 拟议的方法是创新的，因为它以前从未在 提出的方法，这对NIH的使命很重要，因为理解 大脑对美国国立卫生研究院的任务非常重要。许多疾病和残疾 由于大脑功能受损或受损所致。名单很长：药物滥用， 因衰老、耳聋、学习而导致的失明、记忆和认知障碍 儿童残疾、所有类型的精神疾病、创伤后应激 障碍、癫痫、创伤性脑损伤、中风等。如果目标是 修复、预防和治疗这种疾病，从根本上深刻理解 神经元活动意味着什么，这一活动意味着什么 思维过程、感觉、动机和我们影响 世界将从改进的神经信息处理理论中受益。

DESCRIPTION (Adapted from the applicant's abstract): If the purpose of the nervous system is information processing, then we should be able to say what it means for a biological system to process information. Unfortunately, little can be said in a quantitative way. To address this question attempts have intermittently been made over the last 50 years to understand brain function by applying information theory (IT). In recent years, such attempts have become more frequent, with occasional, interesting successes. However, from the viewpoint of the engineering discipline of IT, these attempts have hardly scratched the surface of what IT has to offer, and so far, people have only attempted to incorporate the most well-known and elementary aspects of IT. Indeed, the ideas applied with apparent success - source coding theory - are essentially as old as Shannon's original work. Since that time much has occurred in IT, including several extensions of Shannon's work. Here the investigators advocate the introduction of RD theory and its recent offspring, successive refinement theory. It is the goal of the proposed research to create a demonstration illustrating the applicability and promising superiority of these more sophisticated results of IT. The investigators will translate the ideas of rate-distortion theory and successive refinement theory from the communication literature, where they were developed, to the issues of biological computation. This translation will necessarily be abstract and mathematical. At the same time, however, the investigators will further show people how to apply these insights as well as test several conjectures. Biologically motivated computer simulations of small examples, examples well within the purview of neural network theory and the issues inherent in studying the computational basis of cognition, will be used to illustrate the theory being developed. Finally, the investigators describe how RD theory and the dynamics inherent in the translated version of successive refinement theory can be used to quantify some of the most pervasive metaphors of neural computation. Thus although there is tremendous promise in using the known results of IT, the investigators believe that neuroscientists must begin by applying some of the deeper aspects of the theory. In particular, the overly simplistic uses of IT which now exist in the neuroscientific literature must be clarified and upgraded. The proposed approach is innovative because it has not been done before in the way proposed and it is important to the mission of NIH because understanding the brain is important to the mission of NIH. Many diseases and disabilities result from impaired or damaged brain function. The list is long: drug abuse, blindness, memory and cognitive impairments due to aging, deafness, learning disabilities in children, all types mental illness, post-traumatic stress disorder, epilepsy, traumatic brain injury, stroke, etc. If the goal is to repair, prevent, and treat such maladies, a fundamental, deep understanding of what neuronal activity signifies and what this activity means in terms of thought processes, sensations, motivations, and our ability to affect the world will benefit from an improved theory neural information processing.