Dynamic Neural Network Width Adjustment for Wireless Resource Allocation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current wireless communication network resource allocation methods struggle with dynamic adjustment on demand, particularly in 6G networks with high heterogeneity and dynamics, leading to challenges in optimizing resource allocation for diverse user needs and varying computing resources.
Innovation Solution
A method implemented in a server within a wireless communication network that dynamically adjusts resource allocation on demand using a neural network. This involves obtaining task feature information and CPU frequency, determining a target optimal resource allocation model from a knowledge base, and using this model to allocate bandwidth or power resources to user devices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If iteration-based methods are used to solve resource allocation optimization problems, then optimality of resource allocation is improved, but decision-making time increases significantly
Solution Approach 1:
The patent pre-trains multiple neural network models with different widths offline before actual resource allocation. During runtime, the system selects from these pre-trained models based on current task requirements and CPU frequency, avoiding the need for real-time iteration and optimization. This preliminary preparation resolves the contradiction by shifting computational burden from runtime to training time.
Solution Approach 2:
The patent dynamically selects the appropriate neural network model width based on real-time CPU frequency and task characteristics. The system adjusts the model complexity (width) according to available computing resources and service requirements, enabling flexible trade-off between decision-making speed and allocation optimality without fixed iteration-based approaches.
2Manufacturing precision
If neural network width is increased to improve resource allocation accuracy, then optimality is improved, but computational complexity and decision-making time increase
Solution Approach 1:
The patent changes the parameter of neural network width dynamically based on CPU frequency and task requirements. By adjusting this structural parameter, the system can adapt the computational complexity and accuracy of the model to match available resources and service level agreements, resolving the contradiction between accuracy and complexity.
Solution Approach 2:
The patent segments the neural network into multiple width configurations (different model sizes) that can be selectively deployed. Instead of using a single fixed-width model, the system divides the solution space into multiple discrete model options, each optimized for different computational resource scenarios, allowing selective deployment based on current system state.
3Device complexity
If fixed-width neural network models are used, then device complexity is reduced, but adaptability to different computing resources and service requirements deteriorates
Solution Approach 1:
The patent creates a universal resource allocation system that can handle multiple different computing resource scenarios and service requirements using a family of pre-trained models with different widths. The system selects the appropriate model from this family based on current CPU frequency and task characteristics, achieving multi-functionality without requiring a single complex adaptive model.
4Adaptability or versatility
If larger neural network models are deployed to handle diverse 6G services, then service coverage and adaptability are improved, but resource consumption and decision-making time increase
Solution Approach 1:
The patent dynamically adjusts the neural network model width based on real-time CPU frequency and service requirements. For computationally intensive 6G services requiring high adaptability, larger models are selected. For less demanding services or when CPU frequency is low, smaller models are deployed. This dynamic adaptation resolves the contradiction between service coverage and resource consumption.
Data Source
AI summary
A wireless communication network resource allocation method implemented in a server in a wireless communication network, includes: obtaining task feature information of each user device and a CPU frequency of the server in each time slot; obtaining a task data volume average value; determining, based on a knowledge base including sample data groups and optimal resource allocation models, a target optimal resource allocation model matched with the task data volume average value and the CPU frequency of the server; obtaining, based on the task feature information of the user devices in the time slot and the target optimal resource allocation model, resource allocation results of the user devices, and transmitting task data to the user devices based on the results. A width of a dynamic neural network can be automatically adjusted according to task features and computational capacity, and on-demand adjustment of decision speed and resource optimality can be realized.