PsyNeuLink: A Block Modeling Environment for Cognitive Neuroscience and Computational Psychiatry

PsyNeuLink：认知神经科学和计算精神病学的块建模环境

• 批准号: 9976610
• 负责人: JONATHAN D COHEN
• 金额: $ 19.75万
• 依托单位: PRINCETON UNIVERSITY
• 依托单位国家: 美国
• 财政年份: 2019
• 资助国家: 美国
• 起止时间: 2019-07-15 至 2022-04-30
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/9976610
• 关键词: Acceleration; Address; Architecture; Attention; Basal Ganglia; Biological; Biological Models; Brain; Complement; Complex; Computer Models; Computer software; Custom; Data; Development; Documentation; Education; Educational Curriculum; Environment; Episodic memory; Evaluation; Exercise; Explosion; Foundations; Goals; Grain; Growth; Hippocampus (Brain); Human; Individual; Influentials; Libraries; Literature; Maintenance; Manuals; Mental disorders; Methods; Modeling; Modernization; Neurosciences Research; Perceptual learning; Play; Prefrontal Cortex; Procedures; Psychiatry; Publications; Publishing; Pythons; Research; Role; Seeds; Short-Term Memory; Speed; Statistical Data Interpretation; System; Time; Writing; application programming interface; base; cognitive function; cognitive neuroscience; cognitive process; deep learning; deep neural network; design; flexibility; head-to-head comparison; improved; learning network; memory encoding; memory retrieval; mental function; nervous system disorder; neural model; open source; parallelization; preservation; programs; relating to nervous system; repository; simulation; tool; tool development

Project Summary Paralleling the growth of neuroscience research, there has been an explosion in the development of computationally explicit models of the functions of core brain subsystems. Unfortunately, however, there has not been a commensurate development of the tools needed to share, validate, and compare such models, or integrate them into models of system-level function. Such sharing, evaluation, and integration are necessary if computational modeling efforts are to be useful not only in generating reliable and accurate accounts of how brain subsystems operate, but also of how they interact to give rise to higher cognitive functions, and how disruptions of such interactions may give rise to disturbances of mental function observed in psychiatric and neurological disorders. This proposal seeks to meet this need by developing PsyNeuLink: an open source, Python-based software environment that makes it easy to create new models, import and/or re-implement existing ones, integrate these within a single software environment that will facilitate head-to-head comparison of comparable models, the assembly of complementary models into system-level models, and serve as a common repository for the documentation and dissemination of such models for both research and didactic purposes (i.e., publication, education, etc.). These goals will be pursued under two Specific Aims: 1) Extend the scope of modeling efforts that PsyNeuLink can accommodate by: i) enhancing its application programmer interface (API) used to add new components and interfaces to statistical analysis tools and other modeling environments (such as PyTorch, Emergent and ACT-R; ii) enriching its Library by adding PsyNeuLink implementations of influential models of neural subsystems; and iii) developing a publicly available workbook of simulation exercises as both an introduction to PsyNeuLink and for use in Cognitive Neuroscience and Computational Psychiatry curricula. 2) Accelerate PsyNeuLink by developing a custom compiler that preserves its simplicity and flexibility, while dramatically increasing its speed, to make it suitable for simulation of large and complex system-level models, and for parameter estimation, model fitting, and model comparison. This project will exploit the power and accelerating use of Python, and modern just-in-time compilation methods to develop a tool designed specifically for the needs of systems-level Cognitive Neuroscience and Computational Psychiatry. This promises to open up new opportunities for research at the systems-level — a level of analysis that is crucial both for understanding how human mental function emerges from the interplay among neural subsystems, and how disturbances of individual neural subsystems impact this interplay, disruptions of which are almost certainly a critical factor in neurologic and psychiatric disorders.

项目摘要 随着神经科学研究的发展， 核心大脑子系统功能的计算显式模型。然而不幸的是， 没有相应开发共享、验证和比较这些模型所需的工具，或者 将它们集成到系统级功能模型中。这种共享、评估和整合是必要的， 计算建模工作不仅在产生可靠和准确的账户方面是有用的， 大脑子系统的运作，而且它们如何相互作用，以产生更高的认知功能，以及如何 这种相互作用的中断可能引起精神病学和心理学中观察到的精神功能障碍， 神经系统疾病该提案旨在通过开发PsyNeuLink来满足这一需求：一个开源， 基于Python的软件环境，可轻松创建新模型、导入和/或重新实施 现有的，将这些集成在一个单一的软件环境，这将有助于头对头的比较 可比较的模型，将互补模型组装成系统级模型，并作为 为研究和教学目的记录和传播这些模型的共同资料库 目的（即，出版、教育等）。 这些目标将在两个具体目标下实现：1）扩大建模工作的范围， PsyNeuLink可以通过以下方式进行适应：i）增强其用于添加新内容的应用程序编程接口（API） 统计分析工具和其他建模环境（如PyTorch， Emergent和ACT-R; ii）通过添加PsyNeuLink实现有影响力的 神经子系统;以及iii）开发公开可用的模拟练习工作簿， 介绍PsyNeuLink和用于认知神经科学和计算精神病学课程。（二） 通过开发一个自定义编译器来加速PsyNeuLink，该编译器保留了其简单性和灵活性， 显著提高其速度，使其适合于大型复杂系统级模型的仿真， 以及用于参数估计、模型拟合和模型比较。 该项目将利用Python的强大功能和加速使用，以及现代的即时编译 开发专门针对系统级认知神经科学需求而设计的工具的方法， 计算精神病学这将为系统级的研究开辟新的机会。 这一分析水平对于理解人类心理功能如何从相互作用中产生至关重要， 以及各个神经子系统的干扰如何影响这种相互作用， 其破坏几乎肯定是神经和精神障碍的关键因素。