Distributed AI Media Segmentation for Adaptive Inference Offloading
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems face challenges in efficiently partitioning media data for distributed artificial intelligence (AI) tasks across devices, particularly in determining optimal machine learning models and inference locations, leading to suboptimal processing and resource utilization.
Innovation Solution
A method and device that partition media data into segments, determine machine learning models and parts, and decide whether inference is performed locally or externally, transmitting relevant data and model information for efficient processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If media data is partitioned into segments for distributed AI processing, then processing efficiency and resource utilization are improved, but system complexity and coordination overhead increase
Solution Approach 1:
The patent applies segmentation by dividing media data into multiple segments that can be processed independently by different devices or inference units. Each segment is associated with specific machine learning model parts, allowing distributed processing while maintaining manageable complexity through modular organization of data and computation tasks.
Solution Approach 2:
The patent introduces an intermediary mechanism that coordinates between the first device, inference units, and machine learning model parts. This intermediary layer manages the complexity of distributed processing by handling communication, data flow, and coordination between multiple components, thereby improving processing efficiency without requiring each component to directly manage all system complexity.
2Use of energy by moving object
If machine learning inference is performed externally to the first device, then computational load on the first device is reduced, but communication overhead and latency increase
Solution Approach 1:
The patent applies local quality by allowing different inference operations to be performed in different locations based on their specific requirements. Some machine learning part inferences are performed locally on the first device when speed is critical, while other inferences are performed externally on dedicated inference units when computational resources are needed. This location-dependent approach optimizes the balance between computational load reduction and communication latency.
Solution Approach 2:
The patent implements dynamics by making the inference location flexible and adaptable rather than fixed. The system can dynamically decide whether to perform inference locally or externally based on current conditions such as computational resource availability, data characteristics, and performance requirements. This dynamic approach allows the system to optimize computational load distribution while minimizing communication latency as conditions change.
3Adaptability or versatility
If distributed AI processing is implemented across multiple devices, then resource utilization is improved, but coordination and data management complexity increase
Solution Approach 1:
The patent applies universality by designing a coordinated system where the first device and inference units can perform multiple functions. The first device can both process media data segments and coordinate with external inference units, while inference units can process different machine learning model parts. This multi-functional design improves resource utilization across the distributed system while the standardized coordination protocol manages the complexity of inter-device communication.
Solution Approach 2:
The patent uses parameter changes by dynamically adjusting system configuration parameters such as which machine learning model parts are processed where, how media data is segmented, and the level of coordination required. By changing these parameters based on system conditions, the patent optimizes resource utilization across distributed devices while adapting the coordination complexity to match current operational requirements.
Data Source
AI summary
Procedures, methods, architectures, apparatuses, systems, devices, and computer program products for distributed Artificial Intelligence, AI. A first device partitions a unit of media data corresponding to a media content item into a plurality of media segments, and, for each media segment, determines a machine learning model, a machine learning part, and whether inference using the machine learning part is to be performed externally to the first device or by an inference unit of the first device, transmits content data corresponding to the media segment and information indicative of the determined machine learning model and machine learning part for inference, and obtains processed data resulting from processing of the content data using at least the machine learning part.