ML Model Configuration via Data Characteristic Comparison

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

The conventional process of configuring and training machine learning models is time-consuming and computationally expensive due to the need for iterative experimentation with various model types and parameters to determine optimal settings.

Innovation Solution

A machine learning model configuration framework that determines optimal configuration and training parameters based on data characteristics of prospective input values, eliminating the need for iterative experimentation by comparing and clustering data characteristics with previously built models to apply similar parameters.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If iterative experimentation with various model types and parameters is performed to determine optimal settings, then model performance accuracy is improved, but computation processing time and memory usage increase significantly

Engineering Contradiction:

Improvemodel performance accuracyVSAvoidconfiguration and training time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The system performs preliminary analysis of data characteristics (distribution, type, dimensionality) before model configuration. By pre-characterizing the input data and comparing it against stored profiles of previously successful configurations, the system determines optimal model parameters without requiring iterative experimentation, thus resolving the time-accuracy tradeoff

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system creates a library of previously successful model configurations with their associated data characteristics. When a new model needs to be configured, the system copies and adapts parameters from similar historical configurations rather than re-experimenting, enabling rapid setup while maintaining accuracy through learned patterns

Inventive Principle:

Principle #26Copying

2Measurement precision

If iterative experimentation with various model types and parameters is performed to determine optimal settings, then model performance accuracy is improved, but computer processing power consumption increases

Engineering Contradiction:

Improvemodel performance accuracyVSAvoidcomputer processing power consumption

Core Design Contradiction:

Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The system performs preliminary characterization of data features and compares them against pre-stored successful configuration profiles. This pre-computation approach determines optimal parameters through pattern matching rather than iterative training experiments, significantly reducing CPU/GPU processing power requirements while maintaining model accuracy

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system replaces the mechanical iterative trial-and-error process with an information-based approach using data characteristic comparison and pattern recognition. By substituting computational experimentation with analytical comparison of data profiles, the system reduces processing power consumption while achieving the same configuration goal

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

3Measurement precision

If iterative experimentation with various model types and parameters is performed to determine optimal settings, then model performance accuracy is improved, but memory usage increases

Engineering Contradiction:

Improvemodel performance accuracyVSAvoidmemory usage

Core Design Contradiction:

Measurement precisionVSQuantity of substance

Solution Approach 1:

The system extracts and stores only the essential data characteristics (distribution type, dimensionality, feature types) rather than storing complete training datasets or multiple model versions. This selective extraction reduces memory requirements while preserving the information needed to determine optimal configurations through comparison

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The system creates a compact library of configuration profiles containing summarized data characteristics and successful parameter sets. By copying only the essential configuration metadata rather than full model artifacts, the system maintains reference information for accurate configuration while minimizing memory consumption

Inventive Principle:

Principle #26Copying

Data Source

PatentUS20240386317A1Systems and methods for configuring and training machine learning models based on input value characteristics

Publication Date: 2024.11.21 PAYPAL INC

AI summary

There are provided systems and methods for configuring and training a machine learning model based on data characteristics associated with input data usable for the machine learning model. First measures are derived for the machine learning model based the input data usable for the machine learning model. The first measures represent data characteristics associated with the input data. The first measures are compared against measures associated with other previously built machine learning models. Based on a comparison, a particular previously built machine learning model having data characteristics most similar to the data characteristics calculated for the machine learning model is selected. A machine learning model configuration setting and training parameters may be determined based on the particular previously built machine learning model. The machine learning model is configured and trained based on the configuration setting and training parameters.