CRCNS: The neural basis of probabilistic inference in the visual system

将传入的感官信息与先前学到的知识相结合是基本的 对于感觉皮质的计算，人们仍然知之甚少。为了研究精神分裂症的神经基础 在视觉皮层的计算中，我们结合了三个关键技术。首先，我们聘请了一名严格的 用数学框架来预测感觉神经元的活动应该如何随着 学习和依赖于任务。其次，我们通过消除这些因素来随意操控神经回路 被假设为携带先前学习的信息的输入。第三，我们记录扣球 在有和没有这些输入的情况下，许多初级视觉皮质(V1)神经元同时活动。 我们将在三个重要场景中结合使用这三种技术。在第一个场景中，我们将测量 并分析V1响应在学习两个不同版本的 方位辨别任务。我们将使用这些新数据来验证我们的理论框架并进行比较 它指向另一种理论。在第二个场景中，我们将分析V1响应，而主题是 多任务处理，在两个不同的任务之间切换。这将提供对性能来源的洞察 由于多任务处理的局限性，引入了层次化决策的基础。在第三种情况下，我们 \‘{i_on1p are V1活动以进行顺序判断’.rnakingtaskswh！Jn the Brain Over-weig_h_ts e？rly E)Vi_Dence) (显示“确认偏差”)，以及何时不显示。这将使我们能够测试一种新的计算 对视觉领域的确认偏差进行了解释。 我们的结果将解决系统神经科学中的几个重要争论： 感觉区域的反馈连接？感觉相关变异性的来源和作用是什么？ 回应是什么？哪个数学框架能最好地描述感觉计算？ 相关性(请参阅说明)： 这个项目将研究大脑如何将视网膜上的视觉信息与先验知识结合起来。 关于世界，以支持视觉感知和决策。对神经学的洞察 这些过程的基础将帮助我们了解诸如精神分裂症、自闭症和 视觉处理和功能方面的ADHD。

Combining incoming sensory information with previously learned knowledge is one of the fundamental computations of sensory cortex, yet It remains poorly understood. In order to investigate the neural basis of this computation in visual cortex we combine three key techniques. First, we employ a rigorous mathematical framework to make predictions about how the activity of sensory neurons should change with learning and depend on the task. Second, we causally manipulate the neural circuitry by eliminating those inputs that have been hypothesized to carry previously learned information. Third, we record the spiking activity of many primary visual cortex (V1) neurons simultaneously, with and without those inputs. We will combine these three techniques in three important scenarios. In the first scenario, we will measure and analyze how V1 responses change over the course of learning two different versions of an orientation-discrimination task. We will use this new data to validate our theoretical framework and compare it to alternative theories. In the second scenario, we will analyze V1 responses while the subject is multitasking, switching between two different tasks. This will provide insights into the source of performance limitations due to multitasking and into the basis of hierarchical decision-making. In the third scenario, we \'{ill c_on1pare V1 activity for sequential de~ision'.rnaking taskswh!Jn the brain over-weig_h_ts e§rly E)Vi_dencE) (displaying a 'confirmation bias'), and when it does not. This will allow us to test a new computational account of the confirmation bias in the visual domain. Our results will address several important debates in systems neuroscience: What is the function of feedback connections to sensory areas? What is the source and role of correlated variability of sensory responses? What mathematical framework best describes sensory computations? RELEVANCE (See instructions): This project will study how the brain combines the visual information on the retina with prior knowledge about the world, in order to support visual perception and decision-making. Insights into the neurological basis of those processes will help us understand the effect of diseases such as schizophrenia, autism, and ADHD on visual processing and function.