Mobile ML Inference Splitting for Privacy and Device Constraints
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing AI/ML inference processes in mobile devices face challenges due to computation, memory, and energy constraints, necessitating a method to optimally split the process between user equipment (UE) and cloud/edge servers based on privacy, delay, computation, and energy requirements.
Innovation Solution
A method involving user equipment (UE), application function (AF), network endpoints (NEs), network data analytics function (NWDAF), and network exposure function (NEF) to determine how to split the ML inference process by exchanging information and generating decision data for optimal distribution across UE and cloud/edge servers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the complete ML inference process is performed onboard the mobile device, then privacy and delay requirements are met, but computation, memory, and energy constraints are violated
Solution Approach 1:
The patent divides the ML inference process into multiple segments that can be distributed between the mobile device and cloud server. The process is segmented into parts that can be executed locally (for privacy-sensitive operations) and parts that can be offloaded to the cloud (for computation-intensive tasks), thereby resolving the contradiction between maintaining reliability requirements and adhering to device constraints.
Solution Approach 2:
The patent introduces a cloud server as an intermediary to handle certain portions of the ML inference process. This intermediary accepts computational tasks from the mobile device, processes them, and returns results, enabling the system to overcome local device constraints while maintaining privacy and delay requirements through selective offloading.
2Device complexity
If the ML inference process is offloaded to cloud server, then computation and energy constraints are satisfied, but privacy and delay sensitivity are compromised
Solution Approach 1:
The patent applies local quality by executing specific portions of the ML inference process locally on the mobile device rather than offloading everything to the cloud. Privacy-sensitive operations and delay-critical tasks are retained locally, while only computation-intensive portions are offloaded, thereby maintaining reliability requirements while satisfying device constraints.
Solution Approach 2:
The patent segments the ML inference process into distinct executable portions that can be selectively placed on the mobile device or cloud server based on their specific requirements. This segmentation enables fine-grained control over which operations maintain privacy and delay sensitivity and which can be optimized for computational efficiency.
3Productivity
If the ML inference process is split between UE and cloud server, then optimal resource utilization is achieved, but system complexity increases
Solution Approach 1:
The patent designs a multi-functional system where the mobile device can both execute ML inference operations locally and communicate with the cloud server for offloaded processing. The system architecture is designed to handle multiple functions (local execution, cloud offloading, resource management) through unified mechanisms, thereby achieving optimal resource utilization without proportionally increasing system complexity.
Data Source
AI summary
A method performed by a user equipment, (UE), is provided. The method comprises transmitting towards an application function, (AF) a request for splitting an ML inference process. The request comprises any one or more of: information about the UE, information about the ML inference process, and/or a request for information about a network to which the UE is connected. The method further comprises after transmitting the request for splitting the ML inference process, receiving split decision information indicating how to split the ML inference process. The split decision information was transmitted by the AF.