Collaborative ML Recommendation Models for Adaptive Action Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional single predictive models in recommendation systems are limited by accuracy, robustness, and context-sensitivity due to inherent constraints in their structure or parameters, constraining the precision and adaptability of recommendations.
Innovation Solution
A collaborative machine learning model generation system that integrates multiple models trained on different devices to generate personalized recommendations, using one model as a ground truth to train another, thereby improving accuracy and capturing complex relationships between features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional single predictive models are used in recommendation systems, then the system structure remains simple, but the accuracy, robustness, and context-sensitivity of recommendations are limited
Solution Approach 1:
The patent combines multiple predictive models (first predictive model from local device and second predictive model from remote device) into a unified recommendation system. These models are trained on different data sets and perspectives, and their outputs are integrated to generate recommendations that leverage the strengths of each model, thereby improving overall recommendation accuracy while distributing computational complexity across multiple devices
Solution Approach 2:
The patent segments the training process and model execution across different computing devices. The first predictive model is trained locally on the user's device using local data, while the second predictive model is trained remotely on a different device using additional data sources. This segmentation allows each model to specialize in different aspects of user behavior prediction, improving both accuracy and robustness
2Adaptability or versatility
If multiple machine learning models are integrated from different devices, then recommendation accuracy and adaptability improve, but system complexity increases
Solution Approach 1:
The patent creates a universal recommendation system architecture that can accommodate multiple predictive models from different devices while maintaining a consistent interface and training methodology. The system uses a standardized process for model training, data exchange, and recommendation generation that works across diverse device types and data sources, enabling adaptability without proportionally increasing system complexity
Solution Approach 2:
The patent introduces a remote computing device as an intermediary that hosts and manages the second predictive model. This intermediary handles the complexity of remote model training, data synchronization, and result integration, allowing the local device to benefit from enhanced adaptability without directly managing the complexity of multiple distributed models
Data Source
AI summary
A system includes one or more processors to receive a first machine learning model from a first computing device (e.g., a neural network, support vector machine, random forest, etc.) and a second machine learning model from a second computing device; execute the first machine learning model to generate a first recommendation and the second machine learning model to generate a second recommendation; adjust one or more weights or parameters of the second machine learning model; receive a request for one or more potential actions at a first user interface presented on a display of the client device; execute the second machine learning model using an account identifier of a user account being used to access the application; and generate a second user interface on the display of the client device comprising the one or more potential actions.