Probabilistic machine learning and decision making with Monte Carlo particle methods

使用蒙特卡罗粒子方法进行概率机器学习和决策

• 批准号: 249843-2007
• 负责人: deFreitas, Joao(Nando)
• 金额: $ 2.19万
• 依托单位: University of British Columbia
• 依托单位国家: 加拿大
• 项目类别: Discovery Grants Program - Individual
• 财政年份: 2011
• 资助国家: 加拿大
• 起止时间: 2011-01-01 至 2012-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=479365
• 关键词: Probabilistic; machine; learning; decision; making

In 2001, my NSERC proposal began by stating that the modelling of vast amounts of data, such as text, video, multimedia, gene data and environmental signals, was one of the most demanding challenges of artificial intelligence and statistics in the new century. At the time, I pointed out that to construct appropriate models for the data, we needed to contend with uncertainty as data is typically noisy, imprecise and only partially observed. This uncertainty, coupled with the sheer magnitude of the problem, precluded the construction of models using knowledge acquired solely from experts. To overcome these problems, I proposed the use of probabilistic learning techniques and Monte Carlo algorithms to infer the models automatically from data.

2001年，我的NSERC提案一开始就指出，对大量数据（如文本、视频、多媒体、基因数据和环境信号）的建模是新世纪人工智能和统计学面临的最严峻挑战之一。当时，我指出，为了为数据构建适当的模型，我们需要应对不确定性，因为数据通常是有噪的、不精确的并且只能部分观察到。这种不确定性，再加上问题的严重性，排除了利用仅从专家那里获得的知识建立模型的可能性。为了克服这些问题，我提出了使用概率学习技术和蒙特卡罗算法来自动从数据中推断模型。