兔子先生

ABSTRACT

Lockheed Martin Advanced Technology Laboratories developed?a prototype Machine Learning (ML) ecosystem for the AFRL?StreamlinedML program. This ecosystem will significantly improve?the end-to-end process of configuring, training, evaluating, and?deploying machine learning models, and enable ML DevOps for?AF/DoD applications.

?

WHY?

ML model development for AF/DoD applications is often characterized by a narrow set of experts, with domain knowledge and ML?experience, exploring solutions to pressing operational problems in isolation. The model needs an integration ecosystem that democratizes?ML, which will allow wider adoption of ML that lowers the barriers to entry for use.

?

WHAT?

We are building a composable ML ecosystem that includes:

1. ML model?and technology encapsulation using microservices and containerization
2. A runtime framework that enables extensibility through orchestration, including?model-as-a-service (MaaS) capabilities
3. ML workflows to capture ML?abstractions and provide a basis for the runtime composition
4. An integrated?ML repository that allows storage, query, and manipulation of algorithms/models, metadata, (hyper) parameters, and provenance
5. Shared access to?foster a scalable ML (SML) community

?

HOW?

We will develop and deploy a prototype SML?ecosystem that provides SDKs, toolkits, and services to?access and develop best-in-class algorithms by using a?wide variety of ML stacks and datasets. A prototype?version is currently deployed in the AF HPC and AWS?environments; SML users have started evaluating the?APIs and framework.

?