Dynamic ML Model Selection via Data Profiling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Machine learning systems face challenges in maintaining performance with high variance data, as existing techniques like ensemble and cross-validation methods fail to retain unique patterns, leading to diminished model quality and inconsistent operation.
Innovation Solution
The system profiles received data to identify distinct patterns, matching it with stored profiles to dynamically select and apply a suitable machine learning model, ensuring optimal performance even with changing data characteristics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If ensemble and cross-validation approaches are used to improve model quality, then model reliability is improved, but distinct patterns in the underlying data set are lost
Solution Approach 1:
The patent segments the training data into multiple distinct patterns using clustering algorithms. Each pattern is represented as a separate profile with associated machine learning models. This segmentation allows the system to retain and process distinct patterns individually while maintaining overall model reliability through profile-based predictions.
2Device complexity
If a single machine learning model is trained on high variance data, then processing simplicity is maintained, but model performance becomes inconsistent
Solution Approach 1:
The patent implements a dynamic model selection mechanism where the system profiles incoming data and selects the most appropriate machine learning model from multiple pre-trained models based on the data's pattern characteristics. This dynamic adaptation maintains processing simplicity at the inference stage while significantly improving performance consistency through tailored model selection.
3Adaptability or versatility
If multiple distinct patterns are present in training data, then data representativeness is improved, but model training becomes more complex
Solution Approach 1:
The patent performs preliminary action by pre-processing and clustering the training data into distinct patterns before model training. Multiple machine learning models are trained in advance on these segmented patterns and stored as profiles. During inference, the system only needs to profile the incoming data and select the appropriate pre-trained model, significantly reducing the complexity of handling multiple patterns during actual operation.
Data Source
AI summary
Methods, systems and computer readable media are provided for configuring machine learning systems to automatically and dynamically select a machine learning model based on statistical profiling of received data to improve machine learning applications for high variance data. Data is received from a system in operation. A profile is computed for the received data. A database comprising a plurality of stored profiles for a dataset is accessed, wherein each stored profile corresponds to a distinct pattern identified in the dataset. The computed profile is compared to each of the stored profiles to determine whether the computed profile matches one or more of the stored profiles. When one or more stored profiles match the computed profile, a matching profile is selected by the machine learning system. The received data is processed using a ML model associated with the matching profile.