Feature-Insensitive ML Models Using Multi-Source Encoders
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional machine learning models are inflexible regarding input features, requiring reconfiguration and retraining for even minor changes, leading to inefficiencies and performance reductions when data sources become unavailable or new features emerge.
Innovation Solution
A computer modeling system generates encoders to convert features from multiple data sources into a consistent set of intermediate representations, allowing machine learning models to operate independently of data source availability by using these representations as input, thus adapting to changes without retraining.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional machine learning models are configured and trained with a specific set of input features, then the model can perform accurate predictions based on those features, but the model becomes inflexible and requires reconfiguration and retraining when input features change or data sources become unavailable
Solution Approach 1:
The patent segments the input feature processing by introducing an encoder layer that transforms original input features into intermediate representations. This segmentation allows the model to process features from different data sources independently through separate encoders, enabling flexibility when certain data sources become unavailable while maintaining prediction accuracy through the structured intermediate representation layer.
Solution Approach 2:
The patent introduces intermediate representations as an intermediary layer between the original input features and the downstream model processing. This intermediary layer decouples the model from direct dependency on specific input features, allowing the system to adapt to feature changes or data source unavailability without requiring retraining of the entire model while preserving prediction accuracy.
2Adaptability or versatility
If the set of input features for a machine learning model is changed (adding or removing features), then the model can adapt to new data sources or disruptions, but reconfiguration and retraining are required which consume significant resources and time
Solution Approach 1:
The patent performs preliminary action by pre-processing input features through encoders to create intermediate representations before they reach the main model. This preliminary transformation establishes a stable interface that anticipates future feature changes, allowing the system to adapt to new data sources or disruptions without requiring time-consuming reconfiguration or retraining of the entire model.
Solution Approach 2:
The intermediate representations serve as a universal interface that can accommodate multiple different input feature sets and data sources. The encoder layer is designed to handle various input configurations uniformly, enabling the model to process features from different data sources through the same intermediate representation mechanism, thus eliminating the need for separate reconfiguration processes.
3Measurement precision
If a machine learning model relies on specific input features from multiple data sources, then the model can leverage diverse information for better predictions, but the unavailability of certain data sources reduces accuracy or prevents task completion
Solution Approach 1:
The patent implements beforehand cushioning by creating intermediate representations that buffer the model against potential data source unavailability. The encoder layer pre-processes features in a way that cushions the downstream model from direct exposure to missing or disrupted data sources, maintaining prediction accuracy even when certain input features become unavailable.
Solution Approach 2:
By segmenting the feature processing through individual encoders for different data sources, the system can independently handle the availability status of each data source. This segmentation allows the model to maintain reliability by processing available features through their respective encoders while gracefully handling missing data sources without compromising overall prediction accuracy.
Data Source
AI summary
Methods and systems are presented for providing a framework that configures a machine learning model to be insensitive to changes in input features. A computer modeling system determines data sources from which attribute values associated with transactions can be obtained. Instead of configuring the machine learning model to accept the attribute values as inputs, the computer modeling system may configure the machine learning model to accept a vector representation in a multi-dimensional space as input values. The computer modeling system then generates an encoder for each data source. Each encoder is configured to encode attribute values from a corresponding data source to a representation representing the attribute values. Further, each encoder is trained to minimize a variance between outputs of the different encoders. The computer modeling system determines a vector representation based on the representations generated by the encoders and provide the vector representation to the machine learning model.