Hyperparameter Library for Dynamic Synthetic Data Generation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing approaches to generating synthetic data for training artificial intelligence systems are resource-intensive, time-consuming, and costly, requiring the construction of new models for each type of synthetic data needed, which is inefficient and labor-intensive.

Innovation Solution

The implementation of architecture embeddings that map model hyperparameters to data profiles, allowing for the efficient generation of synthetic data by retrieving pre-identified hyperparameter sets from a hyperparameter library based on the desired data profile.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If a new model is constructed for each type of synthetic data needed, then the synthetic data can be generated with required characteristics, but the process becomes time-consuming and resource-intensive

Engineering Contradiction:

Improveability to generate different types of synthetic dataVSAvoidtime required for model construction

Core Design Contradiction:

Adaptability or versatilityVSLoss of time

Solution Approach 1:

The patent pre-constructs and stores multiple synthetic data models with different hyperparameter configurations in a library before they are needed. When synthetic data is required, the system retrieves pre-built models from the library rather than constructing new ones, significantly reducing the time and computational resources needed for each data generation task while maintaining the ability to produce diverse data types.

Inventive Principle:

Principle #10Preliminary action

2Adaptability or versatility

If a new model is constructed for each type of synthetic data needed, then the synthetic data can be generated with required characteristics, but the process becomes costly

Engineering Contradiction:

Improveability to generate different types of synthetic dataVSAvoidcomputational cost

Core Design Contradiction:

Adaptability or versatilityVSUse of energy by moving object

Solution Approach 1:

Inventive Principle:

Principle #10Preliminary action

3Manufacturing precision

If hyperparameter search is performed for each synthetic data generation request, then optimal hyperparameters can be found, but the process becomes slow and expensive

Engineering Contradiction:

Improvehyperparameter optimizationVSAvoidspeed of synthetic data generation

Core Design Contradiction:

Manufacturing precisionVSProductivity

Solution Approach 1:

The patent pre-performs hyperparameter search and model construction for various data profiles, storing the optimized hyperparameter sets and corresponding models in a library. When a new synthetic data request arrives, the system retrieves pre-optimizedhyperparameters from the library matching the required data profile, eliminating the need for time-consuming hyperparameter search during actual data generation while maintaining high data quality.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent creates and stores copies of optimized models and hyperparameter configurations in a library. Instead of performing expensive hyperparameter optimization from scratch for each data generation request, the system retrieves and reuses pre-optimized model copies that match the required data characteristics, significantly improving generation speed while maintaining model performance.

Inventive Principle:

Principle #26Copying

Data Source

PatentUS12332850B2Systems and methods for architecture embeddings for efficient dynamic synthetic data generation

Publication Date: 2025.06.17 CAPITAL ONE SERVICES LLC

AI summary

Systems and methods for architecture embeddings for efficient dynamic synthetic data generation are disclosed. The disclosed systems and methods may include a system for generating synthetic data configured to perform operations. The operations may include retrieving a set of rules associated with a first data profile and generating, by executing a hyperparameter search, a plurality of hyperparameter sets for generative adversarial networks (GANs) that satisfy the set of rules. The operations may include generating mappings between the hyperparameter sets and the first data profile and storing the mappings in a hyperparameter library. The operations may include receiving a request for synthetic data, the request indicating a second data profile and selecting, from the mappings in the hyperparameter library, a hyperparameter set mapped to the second data profile. The operations may include building a GAN using the selected hyperparameter set and generating, using the GAN, a synthetic data set.