Workshop on Geometry for Signal Processing and Machine Learning

该提案寻求NSF对将于10月13日举行的为期三天的信号处理和机器学习几何研讨会的支持。2016年5月15日，科罗拉多州埃斯蒂斯公园。研讨会由科罗拉多州立大学的Ali Pezeshki（提案负责人）、亚利桑那州立大学的Doug Cochran和科罗拉多州立大学的Louis Scharf组织。研讨会将探讨推动流形建模的应用，并开发一个解决流形推理问题的框架。研讨会的具体目标包括？识别子空间方法范围之外的信号处理和机器学习的挑战性问题，？从现代几何中找出解决这些问题的模型和方法，？确定一组开放的研究问题，将通知NSF？S在信号处理和机器学习方面的持续和未来投资。长期目标是发展一种像现代统计信号处理一样范围广泛、方法精确的理论。这种理论将扩展线性子空间模型，并用几何推理增强统计推理。

This proposal seeks NSF support for a three-day workshop on Geometry for Signal Processing and Machine Learning to be held October 13?15, 2016 in Estes Park, Colorado. The workshop is being organized by Ali Pezeshki, Colorado State University (PI of this proposal), Doug Cochran, Arizona State University, and Louis Scharf, Colorado State University. The workshop will explore applications that impel manifold modeling, and develop a framework for solving inference problems on manifolds. Specific objectives of the workshop include? identifying challenging problems of signal processing and machine learning that lie outside the scope of subspace methods, ? identifying models and methods from modern geometry that address these problems, ? identifying a set of open research questions that would inform NSF?s continuing and future investments in signal processing and machine learning. The long-term objective is to develop a theory as broad in its scope and as precise in its methodologies as modern statistical signal processing. Such a theory would extend linear subspace models and augment statistical reasoning with geometrical reasoning.