ML Input Routing Across Device and Network Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing cellular communication devices face challenges in efficiently routing inputs to machine learned models across different locations while balancing accuracy, latency, privacy, and resource consumption, particularly in scenarios where network connectivity is unreliable or limited.
Innovation Solution
A routing component on the user equipment determines input characteristics and device capabilities to selectively route inputs to machine learned models located on the UE, within the core network, or outside the core network, optimizing for accuracy, latency, and resource usage through parallel processing and model selection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If inputs are routed to machine learned models remotely (in core network or outside core network), then model accuracy and processing capabilities are improved, but network connectivity requirements increase and latency increases
Solution Approach 1:
The system dynamically determines the routing location for machine learned models based on real-time conditions including network connectivity status, device capabilities, and input characteristics. The routing component can select between executing models locally on the device, in the core network, or outside the core network, adapting the execution location to current system state to balance accuracy requirements with latency constraints
Solution Approach 2:
The system changes the execution location parameter of machine learned models based on varying conditions. By monitoring network connectivity, device resources, and model requirements, the system transitions models between different execution locations (local device, core network, external network) to optimize the trade-off between accuracy and response time
2Loss of time
If machine learned models are executed on the device locally, then latency is reduced and privacy is improved, but device resource consumption increases
Solution Approach 1:
The system dynamically adjusts model execution location based on device resource availability and connectivity status. When device resources are sufficient and connectivity is good, models execute locally for low latency. When resources are constrained or network is unavailable, the system routes to remote locations, adapting execution strategy to current device state
Solution Approach 2:
The routing component acts as an intermediary that determines optimal model execution location by evaluating device capabilities, network conditions, and model requirements. It mediates between the application needing model processing and the available execution resources, selecting the best location based on current system state
3Object-affected harmful factors
If inputs are routed outside the core network, then network security is improved by minimizing private data exposure, but network connectivity reliability requirements increase
Solution Approach 1:
The system dynamically selects model execution location based on network security requirements and connectivity reliability. For security-sensitive inputs, the system prefers execution outside the core network when connectivity is reliable. When network reliability is poor, it falls back to core network execution, adapting to current network conditions
Solution Approach 2:
The system changes the execution location parameter based on security and reliability conditions. By monitoring network connectivity status and security requirements, it transitions models between locations to optimize both security and reliability
4Adaptability or versatility
If multiple machine learned models are deployed at different locations, then system versatility and adaptability are improved, but device complexity increases
Solution Approach 1:
The system segments model execution into distinct locations (local device, core network, external network) with clear routing decisions. The routing component handles complexity by making discrete location selection based on evaluated conditions, separating the complexity of model deployment from the simplicity of execution routing
Solution Approach 2:
The routing component provides universal functionality by handling all model routing decisions regardless of location or model type. It serves as a single point of complexity management that handles diverse model deployment scenarios through a unified evaluation and selection process
Data Source
AI summary
Techniques for routing input(s) associated with a machine learned model to various models located at different locations are discussed herein. In some examples, the model may be a generative machine learned model, and in some examples, the different locations may correspond to a first location on a user equipment (UE), a second location in a core network of a network provider, and/or a third location outside of the core network. In some examples, a routing component on the UE may receive an input to a machine learned model and can determine characteristics of the input, and/or characteristics and/or capabilities of the UE, and can route the input to the one or more of the first location, the second location, the third location, or other locations. The UE can receive a response from the location and can present the response at the UE.