Machine Learning and Understanding for | Apply Online

Frequently Asked Questions (FAQs)

1) What is the name of this funding opportunity?

The funding opportunity title is listed as "Machine Learning and Understanding for" (as provided in the opportunity summary).

2) What is the Funding Opportunity Number?

The Funding Opportunity Number is DE-FOA-0001575.

3) Which federal agency and office offered this opportunity?

This opportunity was offered by the U.S. Department of Energy (DOE) Office of Science, through the Office of Advanced Scientific Computing Research (ASCR).

4) What type of program is this (grant vs. contract), and what is the activity category?

This was a discretionary grant program. The activity category was listed as Science and Technology and other Research and Development.

5) What is the CFDA number associated with this opportunity?

The opportunity is associated with CFDA 81.049.

6) What is the main purpose of the opportunity?

The purpose was to support basic research that advances machine learning and "understanding" capabilities specifically tailored for high performance computing (HPC) scientific discovery, especially for extreme-scale (exascale-era) computing platforms and next generation networks.

7) Is the program focused on applying existing machine learning tools to new domains?

No. The emphasis is not on incremental applications of existing tools. The opportunity stresses fundamental advances in algorithms and software designed to operate effectively on extreme-scale computing platforms and next generation networks.

8) What problem is DOE trying to address with this opportunity?

The opportunity highlights that data analysis is becoming a limiting factor for discovery as exascale approaches. Simulations and scientific instruments can produce data faster than scientists can analyze it, and ad hoc workflows often fail to scale. As a result, expensive experimental and simulation outputs may not be fully exploited in a timely manner.

9) What does the opportunity mean by limited "dynamic insight" during execution?

It refers to the difficulty, in many HPC settings, of adjusting an experiment or simulation in real time based on what the data indicates. The opportunity notes that analytics are often not fast, scalable, or integrated enough to support on-the-fly decision-making while a simulation or experiment is running.

10) What does "Machine Learning and Understanding" mean in this context?

In this program framing, it points to a shift toward advanced ML methods combined with systems-level intelligence so that compute-and-data pipelines can become more self-managing and resilient. The "understanding" aspect signals interest beyond black-box prediction, leaning toward approaches that can produce interpretable scientific insight, guide discovery, or support reasoning about complex phenomena at scale.

11) What kinds of capabilities or outcomes is DOE interested in (as described in the summary)?

The summary mentions interests such as automated pattern discovery in large scientific datasets, adaptive analysis that can keep up with data streams, and capabilities that help systems detect issues, recover, and "heal" with reduced manual intervention.

12) What technical realities of HPC/exascale systems were proposals expected to consider?

The opportunity indicates proposals would typically need to account for the realities of exascale systems, including parallelism, heterogeneity, memory and bandwidth constraints, fault tolerance, and the difficulty of moving data around.

13) What layer of the technology stack is this FOA targeting?

The focus is on basic research challenges at the algorithms-and-software layer, where new methods can be integrated into emerging extreme-scale computing stacks and networking architectures.

14) Does DOE want solutions tailored to a single dataset or scientific domain?

No. The outcome DOE was aiming for is broadly applicable capability rather than narrowly tuned solutions that only work for a single dataset or a single scientific domain.

15) What broader DOE mission motivations are connected to this program?

The motivation is tied to DOE national responsibilities in energy, environmental stewardship, and nuclear security, where scientific progress increasingly depends on extracting insight from massive, complex data produced by simulations, experiments, and observational facilities.

16) Who was eligible to apply?

Eligibility was listed as unrestricted (open to any type of entity), subject to any additional clarifications contained in the full announcement.

17) When was the opportunity posted and created?

The opportunity was posted on April 18, 2016, and created the same day (April 18, 2016).

18) What was the application closing date?

The application closing date was June 21, 2016.

19) How many awards did DOE anticipate making?

DOE anticipated making about 4 awards.

20) What was the maximum award amount (award ceiling)?

The award ceiling was $300,000.

21) Is the opportunity specifically about extreme-scale computing platforms and next generation networks?

Yes. The summary explicitly emphasizes fundamental advances in algorithms and software that can operate effectively on extreme-scale computing platforms and next generation networks.

22) What is the program trying to improve about scientific workflows?

It aims to reduce reliance on one-off, ad hoc workflows that do not scale well, and to enable more scalable, integrated analytics that can keep pace with large data volumes and support more responsive decision-making during execution.

23) Is fault tolerance and resilience part of the intended research direction?

Yes. The summary describes a vision where systems become more self-managing and resilient, including detection of issues and recovery or "healing" with less manual intervention.

24) Is this program framed as basic research?

Yes. The opportunity repeatedly emphasizes basic research and foundational advances required to make extreme-scale machine learning feasible and useful in HPC environments.