CRCNS: Bayesian inference in spiking sensory neurons

CRCNS：尖峰感觉神经元的贝叶斯推理

基本信息

批准号：
9124841
负责人：
LARRY F HOFFMAN
金额：
$ 17.59万
依托单位：
UNIVERSITY OF CALIFORNIA LOS ANGELES
依托单位国家：
美国
项目类别：
财政年份：
2014
资助国家：
美国
起止时间：
2014-09-15 至 2019-08-31
项目状态：
已结题

来源：
https://reporter.nih.gov/project-details/9124841
关键词：
Acceleration Address Afferent Neurons Algorithms Animals Attitude Bayesian Analysis Behavior Brain Brain Stem Caliber Code Collaborations Computers Conflict (Psychology)Data Differential Equation Elements Equation Exposure to External Canal Family Fostering Friendships Germany Goals Head Head Movements Heterogeneity Human In Situ Individual International Investigation K-Series Research Career Programs Label Laboratories Labyrinth Larva Lead Likelihood Functions Location Maps Measurement Measures Mediating Methods Minority Access Modeling Motor Motor Activity Movement Nerve Nervous system structure Neurons New Zealand Pattern Peripheral Phase Population Principal Investigator Probability Process Property Psychophysics Recording of previous events Recruitment Activity Research Role Rotation Sampling Sampling Biases Semicircular canal structure Sensory Sensory Process Signal Transduction Source Specific qualifier value Spinal Spinal nerve root structure Staging Statistical Distributions Statistical Models Stimulus Stochastic Processes Structure Students Swimming Synapses System Tadpoles Techniques Testing Time Training Update Validation Ventral Roots Vestibular nucleus structure Work Xenopus analog base career computational neuroscience computer framework density discount expectation experience graduate student information processing insight interdisciplinary approach kinematics multidisciplinary novel particle peer programs relating to nervous system research study response sensory system summer research theories two-dimensional undergraduate student

项目摘要

DESCRIPTION (provided by applicant): The Bayesian brain hypothesis asserts that nervous systems in humans and animals transmit and process information as probability distributions, for which there is a growing body of psychophysical evidence. However, few investigations have endeavored to investigate Bayesian inference in early stages of sensory processing. Through this investigation we propose to implement such a test in primary afferent and secondary neurons of the inner ear vestibular system. We will explore afferent circuits of the horizontal semicircular canal, a model sensory system dedicated to coding and processing head kinematic state. The strength and novelty of our approach is that we advance a testable theory about how head state information may be represented by peripheral and central sensory neurons as spike measurement densities (i.e. SMDs) and spike posterior densities (i.e. SPDs) respectively. Furthermore, we submit the hypothesis that the central representation of the Bayesian posterior of head kinematic state is updated with new sense data (i.e. by primary afferent SMDs) via a neural analog of a particle filter. The particle filter provides the computational framework to tes whether the dynamic discharge of second-order vestibular neurons represent discrete loci of head kinematic state space. These experiments will be conducted in late-stage Xenopus larvae, from which the natural distribution of head kinematic states can be explicitly determined from videographic analysis of free-swimming animals. This natural distribution will be used to derive the dynamic prior within the particle filter model, and also serve as the basis of turntable stimul that can be used in the laboratories in the US and Germany to record evoked discharge from primary and second-order neurons, respectively. We will utilize the fictive swimming signals recordable from spinal ventral roots to modify the natural phase relationship between locomotor and head movement states by presenting head movement stimuli that conflict with the fictive motor efference copy. We hypothesize that such anomalous representations will lead to predictable errors in the central representation of the dynam-ic posterior through the collective SPDs, which would render strong support of a computational framework of Bayesian state estimation in spiking sensory neurons. Intellectual Merit: The intellectual merit of this proposal is harbored in the direct testing of Bayesian inference in nervous systems through direct neurophysiologic methods. Our goal is to test whether the Bayesian particle filter model extends well beyond just another way of describing spatial and temporal patterns of activity in vestibular neurons. Rather, we posit that i can predict and explain them, thereby advancing a compelling neurocomputational model of Bayesian inference using natural neuronal components. These experiments will also provide new insights into the role of dynamic heterogeneity among vestibular afferent neurons, which will undoubtedly lead to new research strategies that will ameliorate our understanding of vestibular sensory coding. Broader Impacts: The research encompasses broader impacts that include the integrative electrophysiologic and computational neuroscience training for individuals at the postdoctoral, graduate, and undergraduate levels. The postdoctoral scholar and graduate students will receive an intensely multidisciplinary experiences, with extraordinary opportunities for international collaboration with peers in the US, Germany, and New Zealand. They will receive rigorous exposure to the theoretical and computational aspects of this project. Undergraduate students will be recruited from across the nation as summer research fellows of the Minority Access to Research Careers program. These students, as well as other undergraduate associates participating during the course of the academic year, will have opportunities to engage in laboratory work addressing the multidisciplinary approaches involved in this project. We propose that this will have a significant impact in broadening their perspective of neuroscientific investigation. We have implemented a plan that will enable direct assessments of the undergraduate students' activities in the project, as well as assessments of the research program's impact upon their attitudes and outlook toward creative scientific endeavors and careers. Our goal is that the integrated approach and international friendships fostered by this project will positively impact their confidence and perspective concerning their own scientific creative capabilities.

描述（由申请人提供）：贝叶斯大脑假说断言人类和动物的神经系统以概率分布的形式传输和处理信息，对此有越来越多的心理物理学证据。然而，很少有研究致力于研究感官处理早期阶段的贝叶斯推理。通过这项研究，我们建议在内耳前庭系统的初级传入神经元和次级神经元中实施这样的测试。我们将探索水平半规管的传入回路，这是一种致力于编码和处理头部运动状态的模型感觉系统。我们方法的优势和新颖之处在于，我们提出了一种可测试的理论，关于头部状态信息如何由外周和中枢感觉神经元分别表示为尖峰测量密度（即 SMD）和尖峰后验密度（即 SPD）。此外，我们提出假设，头部运动状态的贝叶斯后验的中心表示通过粒子滤波器的神经模拟用新的感知数据（即通过初级传入 SMD）进行更新。粒子滤波器提供了计算框架来测试二阶前庭神经元的动态放电是否代表头部运动状态空间的离散轨迹。这些实验将在晚期爪蟾幼虫中进行，通过对自由游泳动物的视频分析可以明确确定头部运动状态的自然分布。这种自然分布将用于导出粒子滤波器模型中的动态先验，并作为转盘刺激的基础，可在美国和德国的实验室中用于记录初级和二级神经元的诱发放电，分别。我们将利用从脊髓腹侧根可记录的虚构游泳信号，通过呈现与虚构运动输出副本冲突的头部运动刺激来修改运动和头部运动状态之间的自然相位关系。我们假设这种异常表示将导致通过集体 SPD 的动态后验的中心表示出现可预测的错误，这将为尖峰感觉神经元中的贝叶斯状态估计的计算框架提供强有力的支持。智力优点：该提案的智力优点在于通过直接神经生理学方法对神经系统中的贝叶斯推理进行直接测试。我们的目标是测试贝叶斯粒子过滤模型是否远远超出了描述前庭神经元活动的空间和时间模式的另一种方式。相反，我们假设我可以预测和解释它们，从而使用自然神经元组件推进令人信服的贝叶斯推理神经计算模型。这些实验还将为前庭传入神经元之间动态异质性的作用提供新的见解，这无疑将带来新的研究策略，从而改善我们对前庭感觉编码的理解。更广泛的影响：该研究涵盖更广泛的影响，包括对博士后、研究生和本科生进行综合电生理学和计算神经科学培训。博士后学者和研究生将获得丰富的多学科经验，并获得与美国、德国和新西兰同行进行国际合作的绝佳机会。他们将严格接触该项目的理论和计算方面。将从全国各地招募本科生作为少数族裔研究职业计划的夏季研究员。这些学生以及在学年期间参与的其他本科生将有机会参与实验室工作，解决该项目涉及的多学科方法。我们认为这将对拓宽他们的神经科学研究视角产生重大影响。我们已经实施了一项计划，可以直接评估本科生在该项目中的活动，以及评估研究计划对他们对创造性科学努力和职业的态度和前景的影响。我们的目标是，该项目培养的综合方法和国际友谊将积极影响他们对自己的科学创新能力的信心和看法。