Learning in the Human Brain: Anatomy, Physiology and Computation

该项目将关注两个定义明确的系统中的结构和信息流。首先，我们考虑人脑中一个分层组织的系统，该系统从代表动作目标的前额叶区域延伸到组织运动执行细节的皮质运动区。其次，我们考虑小脑电路如何从这个系统中学习，存储使这种信息处理变得熟练、无意识和自动(例如，机器学习中的摊销)的表征。这两个系统形成了皮质-小脑系统--人类大脑中最大和最突出的网络之一。将有三种紧密结合的方法。首先，将通过对大型的、预先存在的静息状态功能磁共振数据集的分析来研究这些系统的详细结构。其次，使用功能核磁共振的实验工作将测试与额叶活动网络的分层性质相关的假设，并测试与学习相关的可塑性的小脑特征。这些方法将主要由拉姆纳尼监督。第三，在弗里斯顿开发的主动推理框架的指导下，将使用计算模型开发理论神经生物学方法来理解分层网络中的信息流和与学习相关的可塑性。我们将根据学生的兴趣和技能来定制博士学位。实验工作、数据分析和计算模型的灵活平衡将确保博士在未来可能无法获得数据的封锁期间仍然可行。学生将驻扎在皇家霍洛威，在那里他们将与拉姆纳尼研究小组的成员互动。将定期访问弗里斯顿和他在伦敦大学学院惠康信托神经成像中心的研究小组。研究设施(RoyalHolloway)将包括一台现场的、专门用于研究的西门子Trio MRI扫描仪、一个得到全面支持的高性能计算集群，以及Ramnani用于行为研究的实验室设施。

The project will focus on structure and information flow in two well-defined systems.First, we consider a hierarchically organised system in the human brain that extendsfrom prefrontal areas that represent action goals, to the cortical motor areas thatorganise the details of movement execution. Second, we consider how cerebellarcircuitry learns from this system, storing representations that enable this informationprocessing to become skilled, unconscious and automatic (c.f., amortization inmachine learning). These two systems form the cortico-cerebellar system - one of thelargest and most prominent networks in the human brain. There will be three closelyintegrated approaches. First, the detailed structure of these systems will beinvestigated through the analysis of large, pre-existing resting-state functional MRIdatasets. Second, experimental work using functional MRI will test hypotheses relatingto the hierarchical nature of frontal lobe action networks, and test for cerebellarsignatures of learning-related plasticity. These approaches will be primarily supervisedby Ramnani. Third, theoretical neurobiological approaches to understandinginformation flow in hierarchical networks and learning-related plasticity will bedeveloped using computational modelling, informed by the active inference frameworkdeveloped by Friston, who will supervise this aspect. We will tailor the PhD to theinterests and skills of the student. The flexible balance of experimental work, dataanalysis, and computational modelling will ensure that the PhD will remain feasibleduring possible future periods of lockdown when data acquisition may not be possible.The student will be based at Royal Holloway, where they will interact with members ofRamnani's research group. There will be periodic visits to Friston and his researchgroup at Wellcome Trust Centre for Neuroimaging, UCL. Research facilities (RoyalHolloway) will include an on-site, research-dedicated Siemens Trio MRI scanner, a fullysupported,high-performance computing cluster, and Ramnani's lab facilities forbehavioural research.