Cascading Meta Learner for ML Model Metadata Integration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Proprietary machine learning models often fail to leverage metadata associated with data, leading to suboptimal performance in processing big data, limiting their ability to provide accurate predictions and insights.
Innovation Solution
A cascading system that converts a traditional machine learning model into a neural network model, extracts metadata using an embedding model, and concatenates feature vectors to train a combined machine learning model, enhancing the performance of the original model by incorporating both data and metadata information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If proprietary machine learning models are used to process big data, then the models can provide predictions and insights, but the models do not leverage metadata which causes suboptimal performance
Solution Approach 1:
The patent combines the original machine learning model with a metadata processing component into an integrated system. The metadata is extracted from data records and fed alongside the original data into the model, merging two information streams (raw data and metadata) to enhance predictive accuracy and eliminate the loss of metadata information.
2Measurement precision
If metadata is extracted and integrated into the machine learning model, then model performance and accuracy are enhanced, but the system complexity increases
Solution Approach 1:
The patent segments the system into distinct functional modules: a metadata extraction component that processes and extracts metadata from data records, and an integrated machine learning model that consumes both original data and extracted metadata. This segmentation allows the complexity of metadata handling to be isolated in a dedicated component while maintaining enhanced prediction accuracy through the combined information flow.
Data Source
AI summary
Systems as described herein may cascade a meta learner to enhance functionalities of machine learning models. A cascading server may convert a first machine learning model to a neural network machine learning model. The cascading server may use an embedding machine learning model to extract metadata from a plurality of data records. The cascading server may generate a first set of feature vectors using the converted neural network machine learning model, and generate a second set of feature vectors using the embedding machine learning model. The cascading sever may concatenate the first set of feature vectors with the second set of feature vectors to generate concatenated feature vectors. Accordingly, the cascading server may train a combined machine learning model using the concatenated feature vectors and the trained combined machine learning model may enhance performance of the first machine learning model.