Distributed AI Model Subsets for Lower Device Management Overhead
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies face challenges in efficiently managing and optimizing the distribution of Artificial Intelligence (AI) models across multiple devices, leading to suboptimal performance and resource utilization in collaborative AI systems.
Innovation Solution
A method and apparatus for managing distributed AI models, where a first device selects and runs a model subset based on information received from a second device, enabling efficient distribution and utilization of AI services across multiple devices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If AI models are distributed across multiple devices, then system scalability and resource utilization improve, but device complexity and management overhead increase
Solution Approach 1:
The AI model is divided into multiple model subsets that can be distributed across different devices. Each device receives and executes only the relevant model subsets needed for its function, enabling scalable system expansion while keeping individual device complexity manageable through selective model deployment.
Solution Approach 2:
The system includes mechanisms where devices provide feedback about model performance, resource consumption, and execution results. This feedback loop enables centralized coordination to optimize model distribution, reducing management overhead by automatically adjusting allocations based on real-time system conditions and performance metrics.
2Adaptability or versatility
If multiple AI models are deployed on devices, then service functionality improves, but resource consumption and processing load increase
Solution Approach 1:
Each device is assigned specific model subsets tailored to its local capabilities and functional requirements. Rather than uniformly deploying all models across all devices, the system optimizes model allocation to match device hardware characteristics, execution environments, and specific service needs, thereby reducing overall resource consumption while maintaining service functionality.
Solution Approach 2:
The system deploys only the necessary model subsets required for current service execution rather than loading all available models into all devices. This partial deployment strategy reduces resource consumption by executing only relevant model computations while maintaining the ability to scale functionality when needed.
3Productivity
If AI models are selectively executed on devices, then processing efficiency improves, but coordination complexity and communication overhead increase
Solution Approach 1:
A centralized coordination system acts as an intermediary between multiple devices and the distributed AI models. This coordinator manages model selection, allocation, and execution orchestration, simplifying the coordination complexity by centralizing decision-making logic and reducing direct peer-to-peer communication overhead between devices.
Solution Approach 2:
The system performs preliminary model selection and allocation decisions before execution begins. The coordination system pre-determines which model subsets should be deployed to which devices based on service requirements and device capabilities, reducing real-time coordination complexity by establishing execution plans in advance.
Data Source
AI summary
Procedures, methods, architectures, apparatuses, systems, devices, and computer program products for distributed Artificial Intelligence, AI. A first device receives from a second device information indicative of a plurality of distributed Artificial Intelligence, AI, models, wherein each model of the plurality of distributed AI models corresponds to a same AI service, transmits, to the second device, a message including information indicative of a selected distributed AI model from among the plurality of distributed AI models, wherein the selected distributed AI model includes a plurality of model subsets making up the AI service, receives information corresponding to a model subset of the selected distributed AI model, and runs the model subset of the selected distributed AI model based on the information corresponding to the model subset.