An official website of the United States government
ML APIs have greatly relieved application developers of the burden to design and train their own neural network models—classifying objects in an image can now be as simple as one line of Python code to call an API. However, these APIs offer the same pre-trained models regardless of how their output is used by different applications. This can be suboptimal as not all ML inference errors can cause application failures, and the distinction between inference errors that can or cannot cause failures varies greatly across applications. To tackle this problem, we first study 77 real-world applications, which collectively use six ML APIs from two providers, to reveal common patterns of how ML API output affects applications' decision processes. Inspired by the findings, we propose ChameleonAPI, an optimization framework for ML APIs, which takes effect without changing the application source code. ChameleonAPI provides application developers with a parser that automatically analyzes the application to produce an abstract of its decision process, which is then used to devise an application-specific loss function that only penalizes API output errors critical to the application. ChameleonAPI uses the loss function to efficiently train a neural network model customized for each application and deploys it to serve API invocations from the respective application via existing interface. Compared to a baseline that selects the best-of-all commercial ML API, we show that ChameleonAPI reduces incorrect application decisions by 43%.
more »
« less
This content will become publicly available on October 27, 2026
A Science Gateway as a Hub for Island Resilience
The Change Hawaii (Change(HI)) project is fundamentally addressing the existential threat of climate change in Hawaii by integrating data and climate science to foster statewide resilience, enhance decision science, and support workforce development in critical fields. A cornerstone of this initiative is the \textbf{Hawaii Climate Data Portal (HCDP)}, which operates as a vital science gateway and data hub \cite. The HCDP's primary objective is to build capacity through advanced data science and artificial intelligence (AI), serving as a robust resource for monitoring, visualizing, and communicating environmental change \cite{longman_hawaii_2024}. Its critical role is highlighted by its extensive provision of climate data and its Application Programming Interface (API), which is instrumental in the development and functionality of diverse decision support tools tailored for various stakeholders across the state. This paper details the HCDP's integration with the Tapis API platform, and its successful application in developing actionable climate science outcomes for Hawaii.
more »
« less
- PAR ID:
- 10652957
- Publisher / Repository:
- Zenodo
- Date Published:
- Subject(s) / Keyword(s):
- Tapis Science Gateway Decision Support
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
The explosion of IoT devices and sensors in recent years has led to a demand for efficiently storing, processing and analyzing time-series data. Geoscience researchers use time-series data stores such as Hydroserver, VOEIS and CHORDS. Many of these tools require a great deal of infrastructure to deploy and expertise to manage and scale. Tapis's (formerly known as Agave) platform as a service provides a way to support researchers in a way that they are not responsible for the infrastructure and can focus on the science. The University of Hawaii (UH) and Texas Advanced Computing Center (TACC) have collaborated to develop a new API integration that combines Tapis with the CHORDS time series data service to support projects at both institutions for storing, annotating and querying time-series data. This new Streams API leverages the strengths of both the Tapis platform and CHORDS service to enable capabilities for supporting time-series data streams not available in either tool alone. These new capabilities may be leveraged by Tapis powered science gateways with needs for handling spatially indexed time-series data-sets for their researchers as they have been at UH and TACC.more » « less
-
Because of increased variability in populations, communities, and ecosystems due to land use and climate change, there is a pressing need to know the future state of ecological systems across space and time. Ecological forecasting is an emerging approach which provides an estimate of the future state of an ecological system with uncertainty, allowing society to preemptively prepare for fluctuations in important ecosystem services. However, forecasts must be effectively designed and communicated to those who need them to make decisions in order to realize their potential for protecting natural resources. In this module, students will explore real ecological forecast visualizations, identify ways to represent uncertainty, make management decisions using forecast visualizations, and learn decision support techniques. Lastly, students customize a forecast visualization for a specific stakeholder's decision needs. The overarching goal of this module is for students to understand how forecasts are connected to decision-making of stakeholders, or the managers, policy-makers, and other members of society who use forecasts to inform decision-making. The A-B-C structure of this module makes it flexible and adaptable to a range of student levels and course structures. This EDI data package contains instructional materials and the files necessary to teach the module. Readers are referred to the Zenodo data package (Woelmer et al. 2022; DOI: 10.5281/zenodo.7074674) for the R Shiny application code needed to run the module locally.more » « less
-
In collaboration with the Center for Microbiome Analysis through Island Knowledge and Investigations (C-MĀIKI), the Hawaii EPSCoR Ike Wai project and the Hawaii Data Science Institute, a new science gateway, the C-MĀIKI gateway, was developed to support modern, interoperable and scalable microbiome data analysis. This gateway provides a web-based interface for accessing high-performance computing resources and storage to enable and support reproducible microbiome data analysis. The C-MĀIKI gateway is accelerating the analysis of microbiome data for Hawaii through ease of use and centralized infrastructure.more » « less
-
null (Ed.)The effective use of data science - the science and technology of extracting value from data - improves, enhances, and strengthens acquisition decision-making and outcomes. Using data science to support decision making is not new to the defense acquisition community; its use by the acquisition workforce has enabled acquisition and thus defense successes for decades. Still, more consistent and expanded application of data science will continue improving acquisition outcomes, and doing so requires coordinated efforts across the defense acquisition system and its related communities and stakeholders. Central to that effort is the development, growth, and sustainment of data science capabilities across the acquisition workforce. At the request of the Under Secretary of Defense for Acquisition and Sustainment, Empowering the Defense Acquisition Workforce to Improve Mission Outcomes Using Data Science assesses how data science can improve acquisition processes and develops a framework for training and educating the defense acquisition workforce to better exploit the application of data science. This report identifies opportunities where data science can improve acquisition processes, the relevant data science skills and capabilities necessary for the acquisition workforce, and relevant models of data science training and education.more » « less
