Hydro-ML: Symposium on Big Data Machine Learning in Hydrology and Water Resources; Pennsylvania, May 25-29, 2020

点击翻译按钮获取中文摘要

Artificial intelligence has the potential to impact every facet of our society. Machine learning, an important component of artificial intelligence, is revolutionizing much of what we do today. Deep learning is a relatively new subset of machine learning that has tremendous potential to improve our capabilities for industrial applications and scientific discovery. The use of artificial intelligence in hydrology, and deep learning in particular, can bring tremendous benefits to society. Hydrologic academics and professionals, however, have not historically taken advantage of artificial intelligence. While there have been some workshops and gathering opportunities for hydrologists interested in deep learning, there have been no dedicated workshops to build a community that can leverage common datasets, methods, and goals. A Machine Learning in Water Workshop (“Hydro-ML”) is proposed to demystify artificial intelligence for a wider audience in hydrology, build dedicated expertise and collaborative potential through training sessions and hackathons. Building a hydrological machine learning community will enable sharing of resources, organization of competitions stimulating an expansion of the field, and encourage collaboration among academics and professionals to solve large challenges. Special attention will be paid to participants' diversity both in terms of gender and race, which has previously been identified as an issue in artificial intelligence and hydrology. The topics covered by the symposium will be defined through solicitations to the community at large, with the intention of stimulating broader impacts in the hydrologic community. The Hydro-ML symposium will build a collaborative community focusing on machine learning in hydrology. It will consist of four types of sessions: research presentations, breakout discussion forums, deep learning tutorials and hackathons, and community-building activities. Specific sessions will discuss forward-looking, overarching questions that will be solicited from the participants prior to the meeting. These symposium activities will provide the foundation to build collective efforts in novel dataset identification, preparations for machine learning-in-hydrology competitions, and discussion of physically-informed machine learning. The symposium will be widely advertised through diverse channels, including hydrology newsletters and mailing lists, targeted communications to underrepresented, non-profit, and industry groups, and publications aimed at the general public. The organizing committee will encourage participation from diverse and underrepresented communities, and welcome those who are new to artificial intelligence. The outcomes from the symposium include collective publications serving as positional statements by the community (including journal papers and white papers), detailed plans for hydrologic machine learning competitions, and technical deep learning tutorials for audience with varied deep learning experience, ranging from newcomers to intermediate users. The symposium’s events will be aimed at building a diverse community that enables advancements in hydrology which will benefit society.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.