Cognitive and Neural Strategies for Latent Feature Inference

项目摘要 我们周围的世界有一个统计结构，我们可以用它来改善我们的选择。学习基础知识 通过识别关键特征（如变化率）来调整和优化我们的结构， 决策策略。然而，学习这些特征需要在多个领域积累证据。 时间尺度：短时间尺度考虑当前决策的明确证据，长时间尺度考虑 支持潜在环境特征推断。在大脑中，证据在时间尺度上的积累 因此，灵活决策所必需的信息应该在以下地区使用上下文记忆， 海马（HC）。这项建议旨在确定人类的认知策略和神经机制 用于跨时间尺度积累证据，以进行自适应决策。使用跨学科的 一种利用计算建模来开发和验证人类行为和人类行为的方法。 通过电生理学实验，我将1）识别人类用来支持的各种决策策略 多时间尺度推理，2）模拟人类认知策略的合理神经机制，以及3）定义 HC在实现多时间尺度推理中的作用。这项工作符合BRAIN倡议的使命， 链接行为和功能以及优先研究领域5（理论和数据分析工具）和6（人类 神经科学）。与我杰出的导师团队一起，他们将理论和实验方面的专业知识结合起来， 工作，指导阶段的补助金将为我提供1）额外的研究技能，无论是静态的 推理模型和神经电路建模，以及2）通过个性化指导的职业发展， 写作和科学沟通训练。科罗拉多大学博尔德分校提供了一个理想的环境 这项工作，与应用数学，心理学和研究所之间的部门众多的资源， 认知科学此外，随着许多在线课程和研讨会的可用性， 也可访问宾夕法尼亚大学和休斯顿大学的共同导师机构。的 独立阶段研究将联合收割机这一额外的培训与我以前的经验， 电生理学和信号处理来研究HC在灵活决策中的作用，分析人类 癫痫患者在执行多时间尺度决策任务时的神经记录 由我在犹他州大学的合作者完成。我的长期目标是建立自己的实验室， 多模式的理论方法，人类行为和人类神经电生理学，以确定 与灵活决策相关的认知和神经策略，以及病理性 对这些过程的干扰。

PROJECT SUMMARY The world around us has a statistical structure that we can use to improve our choices. Learning the underlying structure by identifying key features, such as the rate of change, is useful for adapting and optimizing our decision-making strategies. However, learning these features requires accumulating evidence across multiple timescales: a short timescale that considers explicit evidence for the current decision, and a long timescale that supports latent environmental feature inference. In the brain, evidence accumulation across timescales necessary for flexible decision-making should therefore engage contextual memory in regions such as hippocampus (HC). This proposal aims to identify cognitive strategies and neural mechanisms humans use to accumulate evidence across timescales for adaptive decision-making. Using an interdisciplinary approach that utilizes computational modeling to develop and validate human behavioral and human electrophysiological experiments, I will 1) identify the variety of decision strategies humans use to support multi-timescale inference, 2) model plausible neural mechanisms of human cognitive strategies, and 3) define HC’s role in implementing multi-timescale inference. This work is in line with the BRAIN initiative’s mission to link behavior and function and priority research areas 5 (Theory and Data Analysis tools) and 6 (Human Neuroscience). With my outstanding mentor team, who have combined expertise in theory and experimental work, the mentored phase of this grant will provide me with 1) additional research skills in both static inference models and neural-circuit modeling and 2) career development through personalized mentorship, writing, and scientific communication training. The University of Colorado Boulder offers an ideal environment for this work, with numerous resources between the departments of Applied Math, Psychology and Institute for Cognitive Science. Additionally, with the availability of many programs and seminars online, resources at co-mentor institutions University of Pennsylvania and University of Houston are also accessible. The independent phase research will combine this additional training with my previous experience in human electrophysiology and signal processing to study the role of HC in flexible decision-making, analyzing human neural recordings from epilepsy patients while they perform a multi-timescale decision-making task recorded by my collaborators at University of Utah. My long term goals are to launch my own lab that applies a multimodal approach of theory, human behavior, and human neural electrophysiology to identifying the cognitive and neural strategies associated with flexible decision-making and the impacts that pathological disruptions have on these processes.