Bayesian Visual Perception: Measuring and Coding Prior Knowledge

贝叶斯视觉感知：测量和编码先验知识

• 批准号: 7918123
• 负责人: Ahna Girshick
• 金额: $ 2.61万
• 依托单位: NEW YORK UNIVERSITY
• 依托单位国家: 美国
• 财政年份: 2009
• 资助国家: 美国
• 起止时间: 2009-08-06 至 2011-01-31
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/7918123
• 关键词: Address; Amblyopia; Area; Behavior; Behavior Therapy; Behavioral; Brain; Case Study; Cataract; Code; Communities; Costs and Benefits; Cues; Data; Decision Making; Effectiveness; Environment; Excision; Foundations; Goals; Gold; Image; Jordan; Knowledge; Learning; Link; Literature; Location; Measures; Modeling; Motion; Nervous system structure; Neurons; Neurosciences Research; Operative Surgical Procedures; Patients; Performance; Population; Probability; Property; Psychophysiology; Public Health; Radial; Research; Retinal Degeneration; Simulate; Solutions; Speed; Strabismus; System; Techniques; Testing; Time; Vision; Visual; Visual Perception; Visual impairment; Visual system structure; Walking; Work; behavior influence; expectation; improved; interest; motor control; multisensory; neurophysiology; novel; novel strategies; optic flow; public health relevance; relating to nervous system; retinal prosthesis; statistics; theories

DESCRIPTION (provided by applicant): We propose to improve our understanding of visual perception and the underlying neural computation in a Bayesian framework. Knowledge of how visual perception and neural computation works in normally sighted people is important for identifying, describing, and developing behavioral therapies for the visually impaired. The Bayesian framework for the brain has become a popular way to model perceptual behavior due to its elegant simplicity, effectiveness, and ability to unite a broad range of topics. However, there are fundamental gaps in our understanding of if and how the brain could operate in a Bayesian manner. The Bayesian hypothesis, which we propose to test, is that prior knowledge about the world has been integrated into our nervous systems and is used in behavior. The first aim of this proposal is to use a novel behavioral technique to quantitatively estimate the form of prior expectations used by observers in visual perception tasks. This will be compared to the statistics of the world to assess the extent that the visual system has integrated prior knowledge. The second aim is to determine the ways in which a neural population code could represent and compute with prior distributions, and to link it to our behavioral data. Finally, we discuss how priors could arise and the time courses over which they could adapt. PUBLIC HEALTH RELEVANCE: The proposed understanding of how the brain uses prior knowledge is pertinent to public health because the Bayesian framework stipulates that patients suffering from visual deficits (e.g. amblyopia, strabismus, and retinal degeneration) may rely more on prior knowledge than others. Our proposed research sets a foundation for addressing the question of whether these patients incur changes to their representations of prior knowledge or the way these representations are used. Additionally, we propose to investigate the time course over which prior expectations can be adapted, which could pave the way for behavioral therapies to assist patients exposed to dramatic changes (e.g. cataract removal surgery or retinal prostheses) by exposing them to tailored new environments.

描述（由申请人提供）：我们建议在贝叶斯框架中提高对视觉感知和潜在神经计算的理解。了解视觉感知和神经计算如何在正常视力的人中起作用，对于识别，描述和开发视力障碍的行为疗法很重要。贝叶斯框架的大脑框架已成为模拟感知行为的一种流行方式，因为其优雅的简单性，有效性和结合广泛主题的能力。但是，我们对大脑是否以及如何以贝叶斯方式运作的理解存在根本的差距。我们建议测试的贝叶斯假设是，关于世界的先验知识已融入我们的神经系统中并用于行为。该提案的第一个目的是使用一种新颖的行为技术来定量估计观察者在视觉感知任务中使用的先前期望的形式。这将与世界统计数据进行比较，以评估视觉系统已整合了先验知识的程度。第二个目的是确定神经人口代码可以用先前的分布表示和计算的方式，并将其链接到我们的行为数据。最后，我们讨论了先生的出现以及他们可以适应的时间课程。公共卫生相关性：对大脑如何使用先验知识的拟议理解与公共卫生有关，因为贝叶斯框架表明，患者患有视觉缺陷的患者（例如，弱视，斜视和视网膜变性）可能比其他人更依赖于先验知识。我们提出的研究为解决这些患者的先验知识表示或使用这些表示形式的方式奠定了基础。此外，我们建议调查可以对先前期望进行调整的时间过程，这可能为行为疗法为暴露于急剧变化的患者（例如白内障去除手术或视网膜假体）铺平道路，通过将其暴露于量身定制的新环境中。