Adaptive Layer for ML Model Upgrade Without Retraining
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Upgrading a POS system from an older to a newer version often results in a significant decrease in the accuracy of machine learning models due to differences in data structures and attributes, making it challenging to maintain prediction accuracy without retraining the model, which is time-consuming.
Innovation Solution
An intelligent adaptive algorithm is implemented as an adaptive layer over the existing ML model, which correlates key features between the old and new systems, generates an extended feature set, and evaluates prediction accuracy to select the best performing features, allowing the model to adapt to the new system without retraining.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If an ML model is retrained with new data from the upgraded system, then the model adapts to the new system, but the process is time-consuming and delays system deployment
Solution Approach 1:
The patent introduces an adaptive layer as an intermediary component between the pre-trained ML model and the upgraded POS system. This adaptive layer acts as a mediator that translates and adapts data from the new system's data structures to the format expected by the existing model, eliminating the need for time-consuming retraining while maintaining model adaptability to the upgraded system
Solution Approach 2:
The patent applies preliminary action by pre-training the ML model on historical data from the old system before the upgrade. The model is trained in advance with all necessary features and relationships, so when the system is upgraded, the pre-trained model can be quickly deployed with minimal adaptation, significantly reducing retraining time
2Measurement precision
If the ML model uses original features from the old system, then the model maintains its trained accuracy, but the model cannot process data from the new system with different data structures
Solution Approach 1:
The patent applies parameter changes by dynamically adjusting the feature parameters based on the data source. When processing data from the upgraded system, the adaptive layer transforms the new data structures into the original feature format that the pre-trained model expects, thereby maintaining prediction accuracy while enabling compatibility with the new system's data structures
Solution Approach 2:
The patent introduces another dimension by adding an adaptive layer that operates between the data input and the ML model. This additional layer provides a transformation dimension that converts data from the new system's structure into the format required by the existing model, allowing the model to maintain its original accuracy while processing diverse data sources
3Adaptability or versatility
If a new POS system is deployed with different data structures and attributes, then the system offers improved functionality, but the existing ML model becomes invalid and prediction accuracy decreases
Solution Approach 1:
The adaptive layer serves as an intermediary that bridges the functionality gap between the new POS system and the existing ML model. It translates the new system's data structures and attributes into the format the model expects, allowing the system to leverage improved functionality while maintaining the model's prediction accuracy through feature transformation
Data Source
AI summary
Aspects of the present disclosure provide systems, methods, and computer-readable storage media that support mechanisms for enabling upgrading of an existing machine (ML) model-based system to an upgraded ML model-based system using an intelligent adaptive algorithm over a previously trained ML model, while maintaining a level of prediction accuracy from the previously trained ML model. In aspects, the intelligent adaptive algorithm is implemented as an adaptive layer over the previously trained ML model, and provides a mechanism to leverage the previously trained ML model to allow the upgraded ML model-based system to perform operations using data from the upgraded ML model-based system and taking advantage of the accuracy of the previously trained ML model. As such, the intelligent adaptive algorithm retains the previously trained information, and makes predictions using new data from the upgraded ML model-based system without having to retrain the previously trained ML model with the new data.