Understanding Multi-Layer Learning in a Biological Circuit

关于神经系统学习的研究主要集中在突触可塑性的影响上。 为正在研究的神经元提供直接的输入。学习环境的模型或 然而，复杂的技能依赖于广泛分布的可塑性，可能发生在突触上 远离驱动决策或行动的神经元。这是众所周知的从多层(或“深”) 人工网络，将学习分布在多个层上要强大得多，但 也比在单层学习更难实现。事实上，计算机科学家 已经解决了这些问题，给人工智能带来了革命性的变化，并正在迅速重塑 人类世界。了解大脑是如何解决这些问题的，无疑是 当今神经科学面临的最大挑战。然而，在这些方面取得了进展 缓慢，部分原因是哺乳动物学习和记忆电路的高度复杂， 如海马体和新大脑皮层，一直是研究的主要焦点。这项建议 将综合的实验和理论方法应用于具有独特优势的系统 了解多层网络中的学习。莫米里德的电感觉叶 鱼是一个持续学习过程的场所，它预测和取消自我产生的感觉 投入，以加强对与行为有关的刺激的检测。在此基础上， 我们建议开发一个从细胞生物物理到网络的ELL模型 动力学，目的是解释突触可塑性如何在加工过程中广泛分布 层次和细胞类型带来了学习。为了实现这一目标，我们将利用尖端技术 高分辨率突触连通性标测及神经监测方法 在整个学习过程中的人口活动。这项拟议的研究预计将 对复杂的学习形式是如何在神经电路中实现的有全面的了解。

Work on learning in neural systems has focused largely on the effects of plasticity at synapses that provide direct input to the neurons being studied. Learning a model of the environment or a complex skill, however, relies on plasticity that is widely distributed and may occur at synapses far from the neurons driving decisions or actions. As is well-known from multi-layer (or 'deep') artificial networks, distributing learning over multiple layers is substantially more powerful but also more difficult to implement than learning at a single layer. The fact that computer scientists have solved such problems has revolutionized artificial intelligence and is rapidly reshaping the human world. Understanding how the brain solves such problems is, undoubtedly, one of the biggest challenges facing neuroscience today. However, progress along these lines has been slow, due in part to the high degree of complexity of learning and memory circuits in mammals, such as hippocampus and neocortex, that have been a major focus of research. This proposal applies integrated experimental and theoretical approaches to a system with unique advantages for understanding learning in multi-layer networks. The electrosensory lobe (ELL) of mormyrid fish is the site of a continual learning process that predicts and cancels self-generated sensory input in order to enhance detection of behaviorally-relevant stimuli. Building on this knowledge, we propose to develop a model of the ELL spanning from cellular biophysics to network dynamics with the goal of explaining how synaptic plasticity widely distributed across processing layers and cell types gives rise to learning. To accomplish this, we will leverage cutting-edge approaches for mapping synaptic connectivity at high-resolution and monitoring neural population activity over the entire time course of learning. The proposed research is expected to yield general insight into how sophisticated forms of learning are implemented in neural circuits.