COMPUTATIONAL SUBSTRATES OF ADDICTION AND REWARD

描述：(申请人摘要) 这项提议的广泛目标是建立一个量化的 涉及奖赏的细胞和分子事件之间的算法联系 处理以及这些事件影响的行为。特别是，我们 将重点放在中脑多巴胺的性质和用途上 系统构建并分布到整个大脑的神经结构。 腹侧节段区多巴胺能神经元的活动和 长期以来，黑质一直被认为是奖赏的过程 刺激物。这些神经元将轴突发送到参与 动机和目标导向的行为。这些相同的多巴胺系统 疾病和滥用药物的目标，如海洛因和 可卡因。在警觉的灵长类动物中，实验表明这些物质的输出 神经元编码未来回报的预测和实际回报之间的误差 未来奖励的次数和大小。预测误差为 显然编码为瞬时峰值速率的变化：值 高于基线意味着目前的状态“好于预期”， 低于预期的值表示当前状态比预期的更差，否则为否 不同之处在于“事情与预期一致”。我们已经开发出一种 中脑多巴胺神经元行为的计算模型 一种自适应优化控制方法--时间法 不同之处。该模型符合电生理和电生理模型。 猴子的行为数据，而且还可以证明执行适当的 确定哪些行动能使长期利益最大化的计算功能 奖励。这个项目的长期目标是提供一个计算性的 多巴胺能机制在分子水平上的作用 和细胞水平以及这些机制对行为的影响 在吸毒成瘾中观察到的。将使用三种方法来实现 这个长期目标是：(1)人类行为实验，(2)数学 分析；(3)复杂环境中虚拟生物的计算机模拟 环境。这三种方法将重点放在 灵长类中的多巴胺能系统，并将被用来探测如何正常 面对奖励做出决定的机制可能会被打乱 上瘾。

DESCRIPTION: (Applicant's Abstract) The broad objective of this proposal is to establish a quantitative algorithmic link between cellular and molecular events involved in reward processing and the behaviors that these events influence. In particular, we will focus on the nature and use of information that midbrain dopamine systems construct and distribute to neural structures throughout the brain. The activity of dopamine neurons of the ventral segmental area and substantia nigra has long been identified with the processing of rewarding stimuli. These neurons send axons to brain structures involved in motivation and goal directed behavior. These same dopamine systems are targets for disruption by disease and by drugs of abuse like heroine and cocaine. In alert primates, experiments show that the outputs of these neurons encode errors between the predictions of future rewards and actual times and magnitudes of future rewards. The prediction errors are apparently encoded as changes in the instantaneous spiking rates: values above baseline mean that the current state is 'better than predicted', values below mean that the current state is 'worse than predicted', and no difference means that 'things are as predicted'. We have developed a computational model of the behavior of midbrain dopamine neurons using a method of adaptive optimizing control called the method of temporal differences. The model is consistent with electrophysiological and behavioral data in monkeys, and also provably executes the appropriate computational function of determining which actions maximize long-term rewards. The long term goal of this project is to provide a computational connection between the action of dopaminergic mechanisms at the molecular and cellular level and the consequences of these mechanisms on behaviors observed in drug addiction. Three approaches will be used to accomplish this long term goal: (1) human behavioral experiments, (2) mathematical analysis, and (3) computer simulations of virtual creatures in complex environments. These three methods will focus on the function of dopaminergic systems in the primate and will be used to probe how normal mechanisms for making decisions in the face of rewards can be disrupted in addiction.