Advancing neural modeling methods and technology

描述(由申请人提供)：今天，可以生成单个神经元和神经电路的非常复杂的计算机模型。这使得研究人员能够以前所未有的方式检查神经元的动力学。这些模型对我们在各个层面上推进对人类神经系统的理解起到了重要作用。然而，管理不断增长的模型复杂性是有问题的，因为它很容易超过我们有意义地理解模型所代表的所有错综复杂的能力。实际上，计算机技术的进步已经超过了模型开发和分析方法的进步。我们的长期目标是开发自动化神经模型生成过程所需的方法、技术和基础设施。本研究项目的目标是重点研究模型表征和仿真技术。这个项目的基本原理是，为了利用不断增长的计算硬件能力而量身定做的严格的模型表征过程，能够对实验数据和模型备选方案进行详细的比较，模型复杂性的向前推进是可控的，并且比现在可能的模型开发时间大大缩短。该项目将追求以下三个目标：1.表征神经模型，其中的目标是利用从模型输入到模型输出的变化维度来改进参数估计和参数敏感性分析的过程。2.量化模型和仿真器稳健性，目标是检验模型变化对模型输出的影响。3.推进主流仿真平台技术，目标是开发最优的神经元模型仿真硬件平台。我们期望最终得到的一套分析工具和模拟平台建议将适用于所有类型的神经模型，将增加我们对这些模型的理解，并减少构建模型本身所需的时间。

DESCRIPTION (provided by applicant): Today it is possible to generate very complex computer models of both single neurons and neural circuits. This has allowed researchers to examine the dynamics of neurons in ways not before possible. These models have been instrumental in advancing our understanding of the human nervous system at all levels. However, managing the ever-increasing model complexity has been problematic as it can easily outstrip our ability to meaningfully comprehend all of the intricacies the model represents. In effect, the advancements in computer technology have out-paced the advancements in the methods by which models are developed and analyzed. Our long-term goal is the development of the methods, technology, and infrastructure necessary to automate the neural model generation process. The objective of this research project is to focus on model characterization and simulation technology. The rationale for this project is that a rigorous model characterization process tailored to exploit the ever-increasing power of computing hardware enables detailed comparisons of experimental data and model alternatives, manageable forward progression of model complexity, and dramatically shorter model development times than are now possible. The project will pursue the following three aims: 1. Characterize neural models, where the objective is to exploit the changing dimensionality from model input to model output to improve the process of both parameter estimation and parameter sensitivity analysis. 2. Quantify Model and Simulator Robustness, where the objective is to examine the effect of model variation on model output. 3. Advance mainstream simulation platform technology, where the objective is to develop optimal hardware platforms for simulating neuron models. It is our expectation that the resulting set of analytical tools and simulation platform recommendations will be applicable to all types of neural models, will increase our understanding of these models, and decrease the time necessary to construct the models themselves.