Cerebellar processing of noisy inputs

摘要/项目摘要 众所周知，小脑是大脑感觉-运动处理的关键部分。基本特征 小脑学习机制的特征是很好的。这种学习是对改进饲料的服务- 前瞻预测，这对小脑的大部分都有助于运动的准确性。这项建议 是关于小脑如何应对噪音输入的，是基于大量的初步数据证明的 一种由小脑实现的新计算，使其对噪声输入的反应更具适应性。 我们首先训练小脑对其苔藓的特定模式做出目标大小的反应 光纤输入。通过系统地改变探头苔藓纤维输入的不同，我们看到小脑 不会随着相异度的增加而降低响应幅度。这样做将是非适应性的，因为 之前没有经验证明这些较小的幅度是正确的。相反，小脑减少了 做出反应的可能性随着差异的增加而增加，但当它确实做出反应时，几乎总是如此 产生正确的幅度。我们提供了大量的初步数据，表明这种适应行为并不是 归因于反应系统天生要么全有要么全无。同样的数据表明，这一点 计算是小脑的，实际上，计算在很大程度上是在小脑皮质完成的。中国政府的回应 浦肯野细胞，小脑皮质的唯一输出，也是全有或全不，并跟踪试验中的行为。 试行基础。在三个具体目标中，我们建议确定这种自适应计算的机制。我们会 第一个完整的记录研究，确定了在广泛的条件下， 在小脑皮质的加工与在小脑深层核团的加工。第二，我们将使用 小脑皮质神经元的记录以确定使反应不确定的机制 投入是随机的。特别是，我们将确定 浦肯野细胞的反应和无反应试验，我们将确定新发现的 浦肯野细胞和篮子细胞之间的连接在放大或限制这些差异方面发挥了作用 使回应随机化。最后，我们将结合使用可逆失活技术和 浦肯野细胞的记录以验证新发现的小脑深核侧支的假设 神经元实现一种形式的传出复制反馈，即使在 输入的信息差异太大，以至于很少引起回应。完成这些研究后，将会确定 一种新的小脑计算适应机制与应对噪声或不确定性有关 投入。这种机制可能是神经病理病因学中一个被低估的类别，原因是 这在很大程度上是因为大多数实验方法寻求消除尽可能多的噪声或变异性 有可能。因此，这些研究代表了对大脑机制的一种新颖和创新的方法。 功能和功能障碍。

Abstract/Project Summary The cerebellum is known to be a key part of the brain's sensory-motor processing. The basic features of cerebellar learning mechanisms are well characterized. This learning is the service of improving feed- forward predictions, which for much of the cerebellum contribute to the accuracy of movements. This proposal is about how the cerebellum copes with noisy inputs and is based on extensive preliminary data demonstrating a novel computation implemented by the cerebellum that makes its responses to noisy inputs more adaptive. We begin by training the cerebellum to output a response of a target size to a particular pattern of its mossy fiber inputs. By systematically varying the dissimilarity of probe mossy fiber inputs we see that the cerebellum does not decrease response amplitude as the dissimilarity increases. Doing so would be non-adaptive, as there is no prior experience that these smaller amplitudes are correct. Instead, the cerebellum decreases the likelihood of making a response as the dissimilarity increases, but when it does respond it almost always produces the correct amplitude. We present abundant preliminary data that this adaptive behavior is not attributable to the response system being inherently all-or-none. The same data demonstrate that this computation is cerebellar and is, indeed, largely accomplished in the cerebellar cortex. The responses of Purkinje cells, the sole output of the cerebellar cortex, are also all-or-none and track the behavior on a trial-by- trial basis. In three specific aims, we propose to identify the mechanisms of this adaptive computation. We will first complete recording studies that identify over a wide range of conditions the relative contributions of processing in the cerebellar cortex versus processing in the cerebellar deep nuclei. Second, we will use recordings from cerebellar cortex neurons to determine the mechanisms that make responding to uncertain inputs stochastic. In particular, we will determine whether there are significant differences in inputs to the Purkinje cells on response versus non-response trials and we will determine whether newly discovered connectivity between Purkinje cells and basket cells play a role in amplifying or thresholding these differences to make responding stochastic. Finally, we will use a combination of reversible inactivation techniques and recordings of Purkinje cells to test the hypothesis that newly discovered collaterals of deep cerebellar nucleus neurons implement a form of efference copy feedback that enforce the proper response amplitude even for inputs that are so different that they rarely elicit a response. Completion of these studies will identify the mechanisms of a novel computational adaptation of the cerebellum related to coping with noisy or uncertain inputs. Such mechanisms may be an under-appreciated category of etiology of neural pathologies, owing largely to the fact that most experimental approaches seek to eliminate as much noise or variability as possible. As such, these studies represent a novel and innovative approach to the mechanisms of brain function and dysfunction.