IEEE 3652.1-2020 pdf free.IEEE Guide for Architectural Framework and Application of Federated Machine Learning.
The infrastructure layer provides capabilities of computing, storage, and communication of long-byte encrypted data units (e.g., 1024 bits) for the operator layer of federated machine learning frameworks. The infrastructure layer supports all the functions needed by traditional machine learning. In addition, it includes the following components and interfaces to support special functions needed by federated machine learning:
A computing component that provides capabilities of long-byte data unit computing operations
A storage component, which provides capabilities of long-byte data unit storing operations to store the encrypted output model and intermediate data, and encrypted data import and export operations, etc.
A communication component that provides capabilities of long-byte data unit communication operations. It supports a large magnitude of dab communication workload (e.g., I or 2 magnitudes higher than traditional distributed ML), and both iterative inter-datacenter and intra-datacenter communication
Interfaces between each component, which provide capabilities of long-byte encrypted data unit transmission and short-byte plain data unit (e.g.. X bits) transmission
Computing components include functions of the following:
Providing services of long-byte data unit computing operations in federated machine learning.
Providing a clear FML task meta-information management scheme to monitor the normal running of federated machine learning activities. The FML-task meta-information management scheme includes FML-scrvice meta computing infomiation management. FML-participant meta computing information management and FML-model meta computing information management. etc.
Providing iPls for developing federated machine learning algorithms in computing aspects. such as providing handles of computing resources for federated machine learning applications to reference and to operate on.
Storage components include functions of the following:
Providing services of long-byte data unit storage operations in federated machine learning. such as long-byte data import and export from other storage systems to storage components.
Having a clear FML-task meta-infonnation management scheme to monitor the normal running of federated machine learning activities.IEEE 3652.1 pdf download.