Computational Model of Motor Sequence Learning

运动序列学习的计算模型

• 批准号: 8322094
• 负责人: F. Gregory Ashby
• 金额: $ 29.51万
• 依托单位: UNIVERSITY OF CALIFORNIA SANTA BARBARA
• 依托单位国家: 美国
• 财政年份: 2003
• 资助国家: 美国
• 起止时间: 2003-06-15 至
• 项目状态: 未结题
• 来源: https://reporter.nih.gov/project-details/8322094
• 关键词: Accounting; Aging; Architecture; Basal Ganglia; Behavior; Behavioral; Benchmarking; Brain Injuries; Brain region; Categories; Computer Simulation; Corpus striatum structure; Cues; Data; Data Reporting; Flavoproteins; Functional Magnetic Resonance Imaging; Globus Pallidus; Goals; Human; Image; Injection of therapeutic agent; Instruction; Learning; Lesion; Mediating; Modeling; Motor; Motor Cortex; Movement; Muscimol; Pathway interactions; Performance; Principal Investigator; Process; Production; Psychological reinforcement; Publishing; Reaction Time; Role; Stroke; Synapses; Synaptic plasticity; System; Testing; Training; Visual; Work; base; novel; programs; relating to nervous system; research study; response; scaffold; sequence learning; therapy design

Project 4 Computational Modeling of Motor Sequence Learning The overall goal of this project is to develop a biologically detailed computational model of learning in the discrete sequence production (DSP) task, which is the task that will be used in all projects of this PPG. The model, which will include multiple regions in premotor and motor cortices, as well as the basal ganglia, will integrate Ashby's work on category learning with Houk's distributed processing module model of the motor system. The final neural architecture used in the model will be based on empirical results from the other PPG projects. Even so, the basic architecture will include extensive cortical-cortical projections, and each cortical region will be connected to the striatum via closed loop pathways. Synaptic plasticity in cortex will be mediated by (2-factor) Hebbian learning, whereas plasticity at cortical-striatal synapses will be mediated by (3-factor) reinforcement learning (RL). Because of this difference, a fundamental hypothesis is that sequence learning in cortex requires initial assistance from the basal ganglia. The key idea is that the basal ganglia input to cortex serves as a critical scaffold for directing cortical plasticity. This approach has been highly successful in explaining category learning and will be extended to sequence learning in the current proposal. Aim 1 is to construct the model and test it against several qualitative benchmarks. These include verifying that the model can learn to make predictive responses (i.e., respond before the next visual cue is presented), and that it can eventually respond without assistance from the basal ganglia. Aim 2 will test the model against some classic published sequence-learning data. The final goal is to test the model that results from completing Aims 1 and 2 against data collected in the other PPG projects. In particular, the goal is for the same basic model to account simultaneously for single-unit recording data collected by Strick and Turner in Projects 1 and 3, for data from their muscimol inactivation experiments, for Strick's flavoprotein imaging data, for fMRI and TMS data collected by Grafton in Project 2, and also for behavioral data collected in all of these projects. Furthermore, a critical goal of the modeling will be to account for changes in all of these data types with the extensive training that is planned in each project. RELEVANCE (See instructions): The proposed work is central to the problem of understanding the mechansims where practice leads to to reorganization of the human motor system in the face of aging, neurodeneration, stroke or brain injury. Understanding these mechansims has an impact on the design of therapies directed at preserving function, developing compensator movements and ultimately, developing novel motor capacity.

项目4运动序列学习的计算建模 这个项目的总体目标是开发一个生物学上详细的学习计算模型， 离散序列生产（DSP）任务，该任务将用于本PPG的所有项目。该模型， 包括运动前区和运动皮质的多个区域，以及基底神经节， 阿什比的工作类别学习与Houk的分布式处理模块模型的运动系统。的 模型中使用的最终神经结构将基于其他PPG项目的经验结果。即便如此， 基本结构将包括广泛的皮质-皮质投射，并且每个皮质区域将被连接 通过闭环通路到达纹状体。皮层突触可塑性将由（2-因子）赫布介导 而皮层-纹状体突触的可塑性将由（3-因子）强化学习介导 （RL）。由于这种差异，一个基本的假设是，皮层中的序列学习需要初始的 来自基底神经节的帮助。关键的想法是，基底神经节输入到皮层作为一个关键的支架 来指导大脑皮层的可塑性这种方法在解释类别学习方面非常成功， 在目前的建议中，将其扩展到序列学习。目标1是构建模型并进行测试， 若干质量基准。其中包括验证模型可以学习做出预测性反应 (i.e.,在下一个视觉提示出现之前做出反应），并且它最终可以在没有帮助的情况下做出反应。 基底神经节Aim 2将针对一些经典的已发表序列学习数据来测试该模型。最终目标是 根据在其他PPG项目中收集的数据，对完成目标1和2后产生的模型进行测试。在 特别是，目标是同一个基本模型同时考虑收集的单个单元记录数据 由Strick和Turner在项目1和3中，用于他们的蝇蕈醇失活实验的数据，用于Strick的 黄素蛋白成像数据，用于Grafton在项目2中收集的fMRI和TMS数据，以及行为数据 在所有这些项目中。此外，建模的一个关键目标将是考虑所有 这些数据类型与每个项目中计划的广泛培训。 相关性（参见说明）： 所提出的工作是理解实践导致的机制问题的核心， 在面对衰老、神经退化、中风或脑损伤时，人类运动系统的重组。 了解这些机制对旨在保留功能的治疗设计有影响， 开发补偿器运动，并最终开发新的电机容量。