17544581. POWER CONTROL IN OVER THE AIR AGGREGATION FOR FEDERATED LEARNING simplified abstract (QUALCOMM Incorporated)
Contents
POWER CONTROL IN OVER THE AIR AGGREGATION FOR FEDERATED LEARNING
Organization Name
Inventor(s)
Eren Balevi of San Diego CA (US)
Taesang Yoo of San Diego CA (US)
Srinivas Yerramalli of San Diego CA (US)
Junyi Li of Fairless Hills PA (US)
Hamed Pezeshki of San Diego CA (US)
POWER CONTROL IN OVER THE AIR AGGREGATION FOR FEDERATED LEARNING - A simplified explanation of the abstract
This abstract first appeared for US patent application 17544581 titled 'POWER CONTROL IN OVER THE AIR AGGREGATION FOR FEDERATED LEARNING
Simplified Explanation
The patent application describes a system for coordinating federated learning among multiple user equipment (UEs) using over-the-air (OTA) aggregation with power control to mitigate aggregation distortion due to amplitude misalignment.
- The system includes a parameter server located at a base station.
- The parameter server selects a group of UEs for an OTA aggregation session based on a common received power property of each UE in the group.
- The parameter server transmits a global model to the selected UEs.
- Each UE in the group trains the global model based on its local dataset and transmits values associated with the trained local model.
- The parameter server receives a combined response from the group of UEs in the form of a first aggregate amplitude modulated analog signal.
Potential applications of this technology:
- Federated learning in wireless communication systems.
- Coordinating machine learning tasks among multiple user equipment.
- Improving the efficiency and accuracy of federated learning in wireless networks.
Problems solved by this technology:
- Aggregation distortion due to amplitude misalignment in federated learning.
- Coordinating and managing federated learning tasks among multiple user equipment.
- Ensuring efficient and accurate training of global models in wireless communication systems.
Benefits of this technology:
- Mitigates aggregation distortion and improves the accuracy of federated learning.
- Enables efficient coordination and management of federated learning tasks.
- Reduces the complexity and resource requirements of federated learning in wireless networks.
Original Abstract Submitted
A parameter server located at a base station may coordinate federated learning among multiple user equipment (UEs) using over-the-air (OTA) aggregation with power control to mitigate aggregation distortion due to amplitude misalignment. The parameter server may select a first group of UEs for a first OTA aggregation session of a federated learning round based on a common received power property of each UE in the first group of UEs. The parameter server may transmit a global model to the first group of UEs. Each UE in the first group may train the global model based on a local dataset and transmit values associated with the trained local model. The parameter server may receive, on resource elements for the first group of UEs, a first aggregate amplitude modulated analog signal representing a combined response from the first group of UEs.
