On-Premise Model Selection via Entity Clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
On-premise machine learning models face challenges due to limited data sets and computational resources, leading to less adapted models and the need for frequent updates through the cloud, which is time and resource-intensive, and raises confidentiality concerns.
Innovation Solution
Generating multiple machine learning models based on clustered telemetric data from similar entities and varying computational complexity, allowing on-premise devices to select appropriate models based on available resources and customer feedback without relying on the cloud for updates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If machine learned models are trained in the cloud using customer data, then model adaptability to specific customers improves, but customer data confidentiality is compromised
Solution Approach 1:
The patent introduces an intermediary mechanism where the cloud-based system generates multiple candidate models using clustered customer data, but the actual model selection and final training occurs locally on the customer's device. This intermediary approach allows the cloud to provide adaptive modeling capabilities without directly accessing or storing sensitive customer data, thus resolving the confidentiality risk while maintaining model adaptability.
2Object-affected harmful factors
If machine learned models are trained locally on on-premise devices, then data confidentiality is preserved, but model performance deteriorates due to limited data and computational resources
Solution Approach 1:
The patent applies preliminary action by pre-generating multiple candidate machine learned models in the cloud using diverse customer data before deployment to the on-premise device. These pre-trained candidate models are then locally evaluated and selected based on performance metrics, allowing the system to leverage cloud computing resources for initial model development while maintaining local data privacy and enabling subsequent local fine-tuning.
3Adaptability or versatility
If cloud-based model updates are performed frequently, then model adaptability improves, but time and computational resources are consumed
Solution Approach 1:
The system performs preliminary generation of multiple candidate models in advance, storing them locally on the on-premise device. When updates are needed, the device can quickly select from pre-generated candidates or perform local fine-tuning without requiring time-consuming cloud communication and retraining, thus reducing update time while maintaining adaptability.
4Adaptability or versatility
If multiple machine learned models are generated and stored locally, then model selection flexibility improves, but device storage and computational complexity increase
Solution Approach 1:
The patent applies partial action by generating and storing only a subset of candidate models locally—specifically, multiple models per cluster category but not all possible models. The system evaluates these partial sets locally and selects the most appropriate model, achieving sufficient flexibility without requiring excessive storage space or computational resources to manage all possible models.
Data Source
AI summary
The disclosed technology relates to a process of providing dynamic machine learning on premise model selection. In particular, a set of machine learned models are generated and provided to an on premise computing device. The machine learned models are generated using a cluster of customer data (e.g. telemetric data) stored on a computing network having different ranges of computational complexity. One of the machine learned models from the set of machine learned models will be selected based on the current available computational resources detected at the on premise computing device. Different machine learned models from the set of machine learned models can then be selected based on changes in the available computational resources and/or customer feedback.