Base Station Model Training Using PDF-Based Representative Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The process of training a base station model using reinforcement learning requires significant computational resources due to the large amount of data needed for preprocessing and simulation parameter finding, which is inefficient and resource-intensive.
Innovation Solution
An electronic device divides base station data into multiple pieces based on time units, superimposes them to generate first data, further divides into smaller intervals, calculates probability density functions for each interval, and generates representative data to train the base station model.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a large amount of base station data is used for preprocessing and finding simulation parameters, then the training of the base station model becomes more accurate, but the computational resource consumption increases significantly
Solution Approach 1:
The patent segments the base station data processing into multiple stages: dividing data into time units, generating first data by superimposition, further dividing into second time intervals, and calculating probability density functions for each interval. This segmentation transforms a single resource-intensive preprocessing step into multiple manageable stages that reduce overall computational burden while maintaining model training accuracy.
Solution Approach 2:
The patent generates representative data that copies the essential statistical characteristics of the original base station data through probability density functions. Instead of using the full raw dataset for model training, the system creates compressed representations that preserve the necessary patterns and distributions, significantly reducing the data volume required for effective model training.
2Productivity
If base station data is preprocessed and simulation parameters are found using traditional methods, then the base station model can be trained, but the process consumes excessive computational resources and time
Solution Approach 1:
The patent performs preliminary actions by pre-calculating probability density functions for different time intervals and generating representative data in advance. This preliminary processing creates a compressed dataset that can be used directly for model training without requiring extensive preprocessing during the actual training phase, thereby reducing both computational resource consumption and time loss.
3Loss of information
If the base station model is trained with raw base station data, then comprehensive information is available for training, but the computational complexity and resource requirements become unmanageable
Solution Approach 1:
The patent transforms the raw base station data into probability density functions, changing the parameter representation from individual data points to statistical distributions. This parameter transformation reduces the dimensionality and complexity of the data while preserving the essential information needed for model training, making the computational process manageable without significant information loss.
Data Source
AI summary
An electronic device includes a memory storing instructions, a transceiver configured to receive base station data, and at least one processor configured to execute the instructions to: divide the base station data into a plurality of pieces of base station data according to a first time unit; generate first data of the first time unit by superimposing the plurality of pieces of base station data on each other; divide the first data of the first time unit into a plurality of second time intervals, according to a second time interval unit; calculate at least one probability density function for each second time interval of the plurality of second time intervals; generate at least one first representative data by using respective probability density functions of the plurality of second time intervals; and train the base station model, based on the at least one first representative data.