AI Training Data Collection Across Wireless Network Nodes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The challenge of clarifying data processing mechanisms for collecting data sets and assistance information for AI model training, especially when collection nodes differ, is urgent in the context of 5G and AI integration in wireless communication.
Innovation Solution
A data processing method and apparatus that involves collecting and sending data sample information, including input-related and label-related information, to a server, with assistance information aiding the training process, and utilizing base stations and core network devices for forwarding these data without processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If data sample information is collected from multiple nodes (terminal, base station, core network device), then the completeness and quality of training data is improved, but the complexity of determining the collection process and coordinating multiple nodes increases
Solution Approach 1:
The patent introduces a server as an intermediary that receives data sample information from multiple nodes (terminal, base station, core network device) and coordinates the data collection process. The server acts as a central hub that manages the aggregation of data from different sources, thereby improving data completeness while avoiding the complexity that would arise from direct multi-node coordination. This intermediary approach allows each node to independently send data to the server without needing to establish complex communication protocols with other nodes.
2Ease of operation
If data sample information is sent to a server for centralized processing, then the convenience of data processing is improved, but the transmission time and network bandwidth consumption increase
Solution Approach 1:
The patent implements preliminary action by enabling terminal devices to perform local preprocessing of data sample information before transmission to the server. The terminal can process and format the data according to required specifications in advance, which reduces the processing burden on the server and accelerates the overall data handling process. This preliminary processing at the source reduces transmission time by ensuring data is ready for immediate processing upon receipt at the server.
3Manufacturing precision
If assistance information is collected and transmitted along with data sample information, then the quality of AI model training is improved, but the quantity of data to be transmitted and processed increases
Solution Approach 1:
The patent applies local quality by selectively collecting and transmitting different types of assistance information based on the specific requirements of the AI model training task. Rather than transmitting all possible assistance information uniformly, the system identifies and transmits only the relevant assistance information needed for particular training scenarios. This approach maintains high training quality by ensuring necessary assistance information is included while reducing the overall data quantity by excluding unnecessary information.
Data Source
Figure 1~4
Figure 5~7
Figure 8~10
AI summary
The present disclosure belongs to the technical field of communications. Provided are a data processing method and apparatus, and a device and a storage medium. The method comprises: collecting data sample information, wherein the data sample information comprises at least one of input-related information and label-related information of an artificial intelligence (AI) model, and the data sample information is used for training the AI model; and sending the data sample information to a server. Provided in the present disclosure is a processing method specific to "data processing", such that when data sample information is collected, the data sample information can be sent to a server, thus reducing the number of situations where a data sample information collection process cannot be determined, and improving the convenience of data processing.