BRAIN EAGER: Integrative Cross-Modal and Cross-Species Brain Models: Motivation and Reward

动机将目标转化为行动，并在几个方面显著影响认知。奖励的动机会使注意力、感知和记忆产生偏差，并增强学习能力，其影响在行为和特定的大脑区域得到证实。然而，考虑到奖励动机的广泛影响，问题仍然存在：奖励动机是如何在整个大脑中传播的，它是如何动态地改变更大的神经回路来引导我们做出适当的行为的？为了回答这些问题，我们开发了动机对神经回路影响的统计模型，该模型结合了不同物种使用不同仪器记录的数据。对动机的神经效应的更全面的理解将阐明它如何影响重要的认知过程，从而产生对从教育到治疗等各个领域的更好表现有用的见解。为了更全面地了解动机的行为和认知效应背后的神经网络动力学，有必要将人类受试者和动物模型的研究结合起来。虽然在人类和动物模型中分别对奖励动机进行了广泛而重要的研究，但显然需要改进跨物种的翻译。由于要解决的问题的复杂性，以及利用每个物种和可用技术提供的优势，一个总体的翻译分析框架是极其重要的。这项工作的目的是开发动态的、分层的贝叶斯模型，以发现可以跨物种和数据收集方式翻译的功能神经网络。人类行为的贝叶斯模型，以及受贝叶斯机器学习启发的神经网络建模方法，已经非常成功，使贝叶斯方法成为探索转化神经网络发现的沃土。

Motivation translates goals into action, and significantly impacts cognition along several dimensions. The motivation for reward biases attention, perception, and memory, and enhances learning, with effects evidenced behaviorally as well as in specific brain regions. However, given such broad effects of reward motivation, the question remains: how does reward motivation propagate throughout the brain and how does it dynamically change the greater neural circuitry to prime us to behave appropriately? In order to answer these questions we develop statistical models for the effect of motivation on neural circuitry, which combine data recorded using different instruments across a variety of species. A fuller understanding of the neural effects of motivation would elucidate how it impacts important cognitive processes, yielding insights that are useful for better performance in various arenas, from education to therapy. In order to gain a more complete understanding of the neural network dynamics underlying the behavioral and cognitive effects of motivation, it is necessary to integrate research in human subjects, and in animal models. While extensive and crucial research has been carried out on reward motivation separately in humans and animal models, there is a clear need for improved translation across species. An overarching analytical framework for translation is extremely important, due to the complexity of the problems being addressed, and to leverage the strengths offered by each species and available technology. The aim of this work is the development of dynamic, hierarchical Bayesian models to discover functional neural networks that can translate across species and data collection modalities. Bayesian models of human behavior, and Bayesian machine learning inspired methods for neural network modeling, have been extremely successful, making Bayesian methods fertile ground for explorations into translational neural network discovery.