Auditory Neuroscience - The architecture and principles governing neural responses to natural sounds

该项目将专注于比较神经对自然声音反应的不同计算模型。最近录音技术的改进，现在允许同时记录数百个神经元对自然声音的反应，必须与能够在单个神经元和网络层面上定量预测和解释这种反应的计算模型相匹配。因此，对新的和现有的建模框架的解释能力进行彻底的比较，对于促进我们对支配听觉大脑的计算原理的理解将是非常宝贵的。这项研究将包括不同的计算单元(例如，线性、非线性、积分)、各种模型体系结构(深层次网络、递归网络、深递归网络)和不同的计算原理(例如，时间预测、预测编码、任务指定)，将编码建模和标准化原理相结合。该项目将使用沿雪貂和老鼠的听觉通路记录的对自然声音的单一单位神经反应进行。该项目的期望是将关于听觉通路的形式和功能的各种想法筛选到最可信的，也希望为未来的听觉大脑组织的实验研究提供参考。

The project will focus on comparing different computational models of neural responses to natural sounds. The recent improvements in recording technology which now allow for simultaneous recording of hundreds of neurons in response to naturalistic sounds must be matched by computational models able to quantitatively predict and interpret such responses, at both the single neuron and network levels. Therefore, a thorough comparison of the explanatory capacity of new and existing modelling frameworks will be invaluable in advancing our understanding of the computational principles governing the auditory brain. This investigation will include different computational units (e.g., linear, nonlinear, integrating), various model architectures (deep hierarchical networks, recurrent networks, deep recurrent networks), and different computational principles (e.g., temporal prediction, predictive coding, task specification), combining encoding modelling and normative principles. The project will be conducted using single-unit neural responses recorded along the auditory pathway of ferrets and mice in response to natural sounds. The expectation is for the project to winnow the large variety of ideas on the form and function of the auditory pathway to only the most plausible, also hoping to inform future experimental investigation of auditory brain organisation