Uncertainty, inference, and introspection in the primate visual system

标题 灵长类视觉系统中的不确定性、推理和自省 摘要 感知系统提供了一扇窗户，可以看到一个不能完全了解的世界。关于世界的不确定性 当感觉提示不完整或相互矛盾时，可以从外部产生，或者在内部，当噪音破坏神经时 申述。理想的感知系统不会忽视不确定性，而是考虑到它。例如，如果一个 感官提示是模棱两可的，先前的经验应该指导对环境的解释。同样， 感官信号的可靠性应该有助于人们对知觉决策的信心。当人类和猴子 在执行知觉任务时，他们通常遵循这些规范性预测。这意味着神经回路 处理感官信息也会调查这些信息的不确定性，并将这一估计用于 感性推理和感性内省。他们是如何做到这一点的还不是很清楚。一个更精确的 了解感觉不确定性的神经加工及其在知觉和认知中的作用可能会有所帮助 推进失认症、自闭症和精神分裂症等病理疾病的治疗。最近，有几个 已经提出了理论框架，为不确定性的神经处理提供了明确的解释 在低级感觉和高级决策电路中(线性概率总体编码，二次 概率总体编码、时间抽样模型和曲线流形假设)。这些 事态发展构成了这项提议的背景。我们将追踪感官不确定性的过程，从 它最初的计算和在早期视觉皮质中的表现，到它最终在告知信心方面的应用 在高水平的知觉决定和调节感觉信号和先验信息的整合中 前额叶皮质的决策区。我们将使用我们收集的数据来比较和对比 各种神经编码理论的预测。首先，我们将研究清醒、凝视中的V1群体活动 猕猴同时呈现其方位不确定性被以不同方式操纵的刺激。接下来，我们将 研究初级视觉皮质中的神经活动如何影响对刺激的知觉估计的信心 定位。最后，我们将确定感官不确定性如何塑造神经群体的表征 在知觉推理任务中的前额叶皮质。这项工作的成果不仅将加强我们的 对视觉系统的理解，也将为研究知觉提供一种新的实验范式 动物的自省和一种新的分析行为自信报告的计算工具。

Title Uncertainty, inference, and introspection in the primate visual system Abstract Perceptual systems offer a window onto a world that cannot be known perfectly. Uncertainty about the world can arise externally, when sensory cues are incomplete or contradictory, or internally, when noise corrupts neural representations. Ideal perceptual systems do not ignore uncertainty, but take it into account. For example, if a sensory cue is ambiguous, prior experience should guide the interpretation of the environment. Likewise, the reliability of sensory signals should inform confidence in perceptual decisions. When humans and monkeys perform perceptual tasks, they often follow these normative predictions. This implies that the neural circuits which process sensory information also survey the uncertainty of this information, and put this estimate to use for perceptual inference and perceptual introspection. How they do so is not well understood. A more precise understanding of the neural processing of sensory uncertainty and its role in perception and cognition may help to advance the treatment of pathologies such as agnosia, autism, and schizophrenia. Recently, several theoretical frameworks have been proposed that offer explicit accounts of the neural processing of uncertainty in low-level sensory and high-level decision-making circuits (linear probabilistic population codes, quadratic probabilistic population codes, temporal sampling models, and the curved manifold hypothesis). These developments form the background for this proposal. We will trace the processing of sensory uncertainty from its initial computation and representation in early visual cortex to its eventual use in informing confidence in perceptual decisions and in regulating the integration of sensory signals and prior information in high-level decision-making areas in prefrontal cortex. We will use the data we collect to compare and contrast the predictions of various theories of neural coding. First, we will study V1 population activity in awake, fixating macaques while presenting stimuli whose orientation uncertainty is manipulated in distinct ways. Next, we will examine how neural activity in primary visual cortex informs confidence in a perceptual estimate of stimulus orientation. Finally, we will identify how sensory uncertainty shapes neural population representations in prefrontal cortex during a perceptual inference task. The outcomes of this work will not only enhance our understanding of the visual system, but will also provide a novel experimental paradigm to study perceptual introspection in animals and a novel computational tool for analyzing behavioral confidence reports.