Analytical Model Tuning via Segmentation and Federated Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing analytical models deployed in different ecosystems, such as cloud, mobile, fog, and edge environments, face compatibility issues due to varying memory and processing requirements, making it challenging to deploy and tune models across these platforms effectively.
Innovation Solution
A method and system for tuning and deploying analytical models by building a global model based on user input data, identifying target ecosystem parameters, selecting a subset of model parameters, updating the model using PMML and custom wrappers, and generating a local model compatible with the target ecosystem, with real-time tuning using federated learning algorithms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a global analytical model is built with comprehensive parameters for cloud environment, then the model achieves high analytical precision, but the model becomes incompatible with edge and mobile environments with limited resources
Solution Approach 1:
The patent segments the global analytical model into multiple local analytical models, each tailored for specific ecosystem parameters. The system divides the comprehensive model into ecosystem-specific versions (cloud, fog, edge, mobile, embedded) by selecting and configuring relevant parameter subsets for each environment, thereby maintaining precision while achieving compatibility.
Solution Approach 2:
The patent applies local quality by customizing each local analytical model with ecosystem-specific parameter configurations. Each environment (cloud, fog, edge, mobile, embedded) receives a model variant optimized for its specific resource constraints and operational characteristics, ensuring that each local model has the appropriate quality and parameter set for its designated ecosystem.
2Measurement precision
If analytical algorithms are implemented with full functionality, then the analytical capability is maximized, but the computation time and resource consumption increase significantly
Solution Approach 1:
The patent implements partial action by deploying only the necessary subset of analytical algorithms and parameters for each ecosystem rather than the full global model. Each local analytical model includes only the algorithms and parameters needed for that specific environment, reducing computation time and resource consumption while maintaining sufficient analytical capability for local decision-making.
Solution Approach 2:
The patent segments the comprehensive analytical algorithm suite into ecosystem-specific algorithm subsets. By dividing the full analytical capability into smaller, environment-appropriate algorithm collections, the system reduces the computational burden on resource-constrained devices while preserving essential analytical functions for each ecosystem tier.
3Productivity
If the analytical model is customized for each ecosystem, then the model achieves optimal performance for that ecosystem, but the complexity of model deployment and management increases
Solution Approach 1:
The patent applies universality by creating a multi-functional model deployment system that can automatically generate and deploy appropriate local analytical models across multiple ecosystems from a single global model definition. The system provides universal model management capabilities that handle customization, deployment, and updates across cloud, fog, edge, mobile, and embedded environments through a unified approach, reducing overall deployment complexity.
Solution Approach 2:
The patent implements preliminary action by pre-configuring the global analytical model with all possible parameters and algorithms, and pre-establishing the framework for generating local model variants. This preliminary setup enables automated model customization and deployment for different ecosystems without requiring complex manual configuration for each environment, thereby reducing deployment complexity while maintaining optimal performance.
Data Source
AI summary
The present disclosure relates to system(s) and method(s) for tuning an analytical model. The system builds a global analytical model based on modelling data received from a user. Further, the system analyses a target eco-system to identify a set of target eco-system parameters. The system further selects a sub-set of model parameters, corresponding to the set of target eco-system parameters, from a set of model parameters. Further, the system generates a local analytical model based on updating the global analytical model, based on the sub-set of model parameters and one or more PMML wrappers. The system further deploys the local analytical model at each node, from a set of nodes, associated with the target eco-system. Further, the system gathers test results from each node based on executing the local analytical model. The system further tunes the sub-set of model parameters associated with the local analytical model using federated learning algorithms.