Collaborative Research: Knowledge Guided Machine Learning: A Framework for Accelerating Scientific Discovery

协作研究：知识引导机器学习：加速科学发现的框架

• 批准号: 1934633
• 负责人: Paul Hanson
• 金额: $ 34.19万
• 依托单位: University of Wisconsin-Madison
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2019
• 资助国家: 美国
• 起止时间: 2019-09-01 至 2022-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1934633&HistoricalAwards=false
• 关键词: Collaborative; Research; Knowledge; Guided; Machine

The success of machine learning (ML) in many applications where large-scale data is available has led to a growing anticipation of similar accomplishments in scientific disciplines. The use of data science is particularly promising in scientific problems involving processes that are not completely understood. However, a purely data-driven approach to modeling a physical process can be problematic. For example, it can create a complex model that is neither generalizable beyond the data on which it was trained nor physically interpretable. This problem becomes worse when there is not enough training data, which is quite common in science and engineering domains. A machine learning model that is grounded by explainable theories stands a better chance at safeguarding against learning spurious patterns from the data that lead to non-generalizable performance. This is especially important when dealing with problems that are critical and associated with high risks (e.g., extreme weather or collapse of an ecosystem). Hence, neither an ML-only nor a scientific knowledge-only approach can be considered sufficient for knowledge discovery in complex scientific and engineering applications. This project is developing novel techniques to explore the continuum between knowledge-based and ML models, where both scientific knowledge and data are integrated synergistically. Such integrated methods have the potential for accelerating discovery in a range of scientific and engineering disciplines. This project will train interdisciplinary scientists who are well versed in such methods and will disseminate results of the project via peer-reviewed publications, open-source software, and a series of workshops to engage the broader scientific community.This project aims to develop a framework that uses the unique capability of data science models to automatically learn patterns and models from data, without ignoring the treasure of accumulated scientific knowledge. Specifically, the project builds the foundations of knowledge-guided machine learning (KGML) by exploring several ways of bringing scientific knowledge and machine learning models together using pilot applications from four domains: aquatic ecodynamics, climate and weather, hydrology, and translational biology. These pilot applications were selected because they are at tipping points where knowledge-guided machine learning can have a transformative effect. KGML has the potential for providing scientists and engineers with new insights into their domains of interest and will require the development of innovative new machine learning approaches and architectures that can incorporate scientific principles. Scientific knowledge, KGML methods, and software developed in this project could potentially be extended to a wide range of scientific applications where mechanistic (also known as process-based) models are used.This project is part of the National Science Foundation's Harnessing the Data Revolution (HDR) Big Idea activity.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.

机器学习（ML）在许多可用数据的应用中的成功导致对科学学科的类似成就的预期日益增长。在涉及未完全理解的过程的科学问题中，数据科学的使用尤其有希望。但是，建模物理过程的纯粹数据驱动方法可能是有问题的。例如，它可以创建一个复杂的模型，该模型既不是训练它的数据，也不是在物理上可以解释的数据之外。当没有足够的培训数据时，这个问题就会变得更糟，这在科学和工程领域很普遍。 以可解释的理论为基础的机器学习模型具有更好的机会来保护从导致不可替代性能的数据中学习虚假模式。当处理至关重要且与高风险有关的问题（例如，生态系统的极端天气或崩溃）时，这一点尤其重要。 因此，在复杂的科学和工程应用中，不可被视为仅限ML或仅科学知识的方法。该项目正在开发新型技术，以探索基于知识的模型和ML模型之间的连续性，在这些模型中，科学知识和数据均协同整合。这种综合方法具有在一系列科学和工程学科中加速发现的潜力。该项目将培训精通这种方法的跨学科科学家，并将通过同行评审的出版物，开源软件和一系列研讨会来传播该项目的结果，以吸引更广泛的科学社区。该项目旨在开发一个框架，以开发一个具有数据科学模型的框架，以自动从数据和模型中学习模型，而无需忽略数据的知识，该项目累积了累积的科学知识。具体而言，该项目通过使用来自四个领域的试验应用将科学知识和机器学习模型探索将科学知识和机器学习模型聚集在一起的几种方法来建立知识引导的机器学习（KGML）的基础：水生态动力学，气候和天气，水文学和转化生物学。选择了这些试点应用，因为它们处于知识引导的机器学习可以产生变革性效果的临界点。 KGML有可能为科学家和工程师提供对其感兴趣领域的新见解，并需要开发创新的新机器学习方法和可以纳入科学原则的建筑。该项目中开发的科学知识，KGML方法和软件可能会扩展到广泛的科学应用，其中使用了机械（也称为基于过程）模型的机械应用程序（也称为基于过程）模型。该项目是国家科学基金会（National Science Foundation）利用数据革命（HDR）的大创意活动的一部分。该奖项是NSF的法定任务，反映了NSF的法规审查，并通过评估范围来进行评估，并具有范围的范围。

项目成果

Modeling dataset: Long-term Change in Metabolism Phenology across North-Temperate Lakes, Wisconsin, USA 1979-2019

建模数据集：1979-2019 年美国威斯康星州北温带湖泊代谢物候学的长期变化

• DOI: 10.6073/pasta/af991c26bace5af8d4b3bb66d7b18af7
• 发表时间: 2022
• 期刊: Environmental Data Initiative
• 作者: Ladwig, Robert;Appling, Alison P.;Delany, Austin;Dugan, Hilary A.;Gao, Qiantong;Lottig, Noah;Stachelek, Joseph;Hanson, Paul C.
• 通讯作者: Hanson, Paul C.

Predicting lake surface water phosphorus dynamics using process-guided machine learning

• DOI: 10.1016/j.ecolmodel.2020.109136
• 发表时间: 2020-08-15
• 期刊: ECOLOGICAL MODELLING
• 影响因子: 3.1
• 作者: Hanson, Paul C.;Stillman, Aviah B.;Kumar, Vipin
• 通讯作者: Kumar, Vipin

Process‐Guided Deep Learning Predictions of Lake Water Temperature

• DOI: 10.1029/2019wr024922
• 发表时间: 2019-11
• 期刊: Water Resources Research
• 影响因子: 5.4
• 作者: J. Read;X. Jia;J. Willard;A. Appling;Jacob Aaron Zwart;S. Oliver;A. Karpatne;Gretchen J. A. Hansen;P. Hanson;William Watkins;M. Steinbach;Vipin Kumar

Data for Lake Mendota Phosphorus Cycling Model

门多塔湖磷循环模型的数据

• DOI: 10.6073/pasta/36d0ee7bf67d9dabade404c92be73917
• 发表时间: 2019
• 期刊: Environmental Data Initiative
• 作者: Hanson, Paul C;Stillman, Aviah B
• 通讯作者: Stillman, Aviah B

Modeled Organic Carbon, Dissolved Oxygen, and Secchi for six Wisconsin Lakes, 1995-2014

1995-2014 年威斯康星州六个湖泊的有机碳、溶解氧和 Secchi 模型

• DOI: 10.6073/pasta/1b5b947999aa2f9e0e95c91782b36ee9
• 发表时间: 2022
• 期刊: Environmental Data Initiative
• 作者: Delany, Austin