Distributed ML DAPP Inference Using Underutilized Edge Compute
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning solutions require significant computational resources, either through on-premise setups which are costly and centralized cloud-computing solutions which have high latency, making them unsuitable for real-time, low-latency, and cost-effective personalized inferences.
Innovation Solution
A distributed machine learning decentralized application (DAPP) platform that leverages underutilized computational power from devices like gaming consoles and autonomous vehicles, distributing inference tasks across a network of service providers to provide real-time, personalized predictions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Power
If traditional machine learning solutions use on-premise computers or centralized cloud-computing, then computational resources are made available, but the cost and latency increase significantly
Solution Approach 1:
The patent segments the centralized cloud-computing system into a distributed network of edge devices. Instead of relying on a single centralized cloud provider, the system divides computational tasks across multiple edge devices (smartphones, tablets, wearables, IoT devices) that are geographically dispersed and closer to the users, thereby reducing latency while maintaining access to computational resources.
Solution Approach 2:
The patent transitions from a vertical hierarchical architecture (client-server-cloud) to a horizontal peer-to-peer mesh network of edge devices. This dimensional shift allows computational resources to be distributed across the network perimeter rather than centralized, enabling parallel processing and reducing the single point of failure and latency associated with centralized cloud access.
2Power
If traditional machine learning solutions use on-premise computers, then computational resources are available, but the cost increases significantly
Solution Approach 1:
The patent enables edge devices to serve dual purposes: they function as both end-user devices and as computational nodes in the distributed machine learning network. This self-service approach allows devices like smartphones and tablets to contribute their processing power to ML tasks, eliminating the need for dedicated expensive on-premise servers and reducing overall system cost.
Solution Approach 2:
The patent makes existing edge devices universal by enabling them to perform both their primary user-facing functions and participate in distributed machine learning computations. This multi-functionality allows a single device to serve multiple purposes, eliminating the need for separate dedicated computational infrastructure and reducing overall system cost.
3Productivity
If a distributed machine learning platform uses underutilized computational power from devices like gaming consoles, then cost-effectiveness and scalability improve, but device complexity increases
Solution Approach 1:
The patent introduces a coordination layer that acts as an intermediary between the distributed edge devices and the machine learning workload. This intermediary manages the complexity of resource allocation, task distribution, and result aggregation, allowing the system to leverage underutilized computational power from diverse devices without exposing the underlying architectural complexity to the end users or application developers.
Data Source
AI summary
A request for an inference from a customer is received at a machine learning (ML) decentralized application (DAPP) platform, where the request includes a data record associated with a user that is associated with the customer. The data record is distributed by the ML DAPP platform among a number of service providers. An inference is received at the ML DAPP platform from each service provider. The received inferences are returned to the customer by the ML DAPP platform.