Multistage Learner for Machine Learning Data Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Large-scale machine learning systems face inefficiencies in processing large datasets due to the need for repeated iterations over extensive training data, where old features dominate the training process, hindering the convergence and accuracy of new features.
Innovation Solution
Implementing a multistage learning approach that focuses on training over subsets of data based on the time features were introduced, allocating more resources to new features and periodically iterating over all data to ensure convergence on both old and new features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If iterative batch learning algorithms process large amounts of training data repeatedly, then model accuracy is improved, but processing time and computational cost increase significantly
Solution Approach 1:
The patent segments the training data into different time-based subsets (first plurality of examples from first time period, second plurality of examples from second time period). This segmentation allows the system to process different data subsets in different iterations, reducing the computational burden of processing entire large datasets repeatedly while maintaining model accuracy through focused learning on relevant data segments.
2Measurement precision
If the machine learning system trains on all available training data, then convergence on old features is achieved, but convergence on new features is hindered due to dominance of old features
Solution Approach 1:
The patent applies local quality by assigning different training priorities to different data subsets based on their temporal characteristics. Examples from more recent time periods (second plurality) are given higher priority in certain iterations, while examples from earlier periods (first plurality) are processed in other iterations. This localized quality adjustment ensures that new features receive adequate attention without being overwhelmed by the volume of older feature data.
3Reliability
If repeated iterations are performed over extensive training data, then model robustness is improved, but processing efficiency decreases
Solution Approach 1:
The patent implements periodic action by alternating between different training iterations that process different data subsets. The system periodically switches between training on the first plurality of examples and the second plurality of examples, creating a rhythmic training pattern that maintains model robustness through multiple iterations while improving processing efficiency by avoiding redundant processing of the entire dataset in each iteration.
Data Source
AI summary
Implementations of the disclosed subject matter provide methods and systems for using a multistage learner for efficiently boosting large datasets in a machine learning system. A method may include obtaining a first plurality of examples for a machine learning system and selecting a first point in time. Next, a second point in time occurring subsequent to the first point in time may be selected. The machine learning system may be trained using m of the first plurality of examples. Each of the m examples may include a feature initially occurring after the second point in time. In addition, the machine learning system may be trained using n of the first plurality of examples, and each of the n examples may include a feature initially occurring after the first point in time.