Gradient-Boosted AI Scaling for Tabular Synthetic Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing generative modeling techniques struggle with scalability and resource efficiency when applied to large-scale, high-dimensional datasets, particularly in domains like healthcare, finance, and telecommunications, failing to handle complex interdependencies and data variability, and lacking robust security measures for sensitive data.
Innovation Solution
A scalable AI model integrating multi-output XGBoost, class-conditional scaling, and dynamic resource allocation, leveraging specialized hardware accelerators like NPUs and TPUs, with encryption and data watermarking to generate high-quality synthetic data that maintains data integrity and confidentiality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If Gradient-Boosted Trees are used for generative modeling on tabular datasets, then model performance is improved, but scalability to larger datasets deteriorates
Solution Approach 1:
The patent segments the GBT model into multiple components including a class-conditional scaler that divides data processing by class, and a multi-output GBT that separates prediction tasks into distinct output channels. This segmentation enables parallel processing and improves scalability while maintaining model performance on large tabular datasets.
Solution Approach 2:
The patent implements dynamic resource allocation where the GBT model adapts its computational resources based on dataset size and complexity. The class-conditional scaler dynamically adjusts scaling parameters per class, and the model structure can be configured to handle varying data dimensions, enabling efficient scaling from small to large datasets.
2Measurement precision
If AI models are trained on large datasets, then data representation accuracy is improved, but computational burden increases
Solution Approach 1:
The patent applies preliminary data scaling and preprocessing through the class-conditional scaler before main model processing. By pre-scaling features according to their class distributions, the model requires fewer computational iterations to converge, reducing overall computational burden while maintaining accurate data representation on large datasets.
Solution Approach 2:
The patent uses synthetic data generation through the GBT-based generative model to create copies of training data. These synthetic copies augment the training dataset, improving data representation accuracy without requiring proportional increases in computational resources for training on actual large datasets.
3Measurement precision
If complex generative models are used to handle data variability, then modeling accuracy is improved, but resource efficiency deteriorates
Solution Approach 1:
The patent creates a multi-output GBT model that serves multiple functions simultaneously: it performs classification, regression, and synthetic data generation within a single unified framework. This multi-functionality maintains high modeling accuracy for complex data variability while improving resource efficiency by avoiding the need for separate specialized models for each task.
4Speed
If traditional scaling methods are applied to tabular data, then processing speed is improved, but handling of data variability deteriorates
Solution Approach 1:
The patent implements class-conditional scaling that applies different scaling transformations to different classes of data rather than a uniform scaling approach. This local quality approach maintains high processing speed through efficient vectorized operations while simultaneously improving adaptability to data variability by tailoring scaling parameters to each class's specific characteristics.
Data Source
AI summary
An example operation may include at least one of loading an Artificial Intelligence (AI) model from the storage, wherein the AI model is one of a diffusion-based model or a flow-based model, receiving tabular input data for execution by the AI model, wherein the tabular input data is scaled with a class-conditional scaler, creating a multi-output Gradient Boosted Tree (GBT), creating a Scalable AI (SAI) model from the AI model by using the multi-output GBT as a function approximator, and generating synthetic data by at least one of: executing the SAI model on the tabular input data or implementing a trained SAI model with the tabular input data, wherein the generating synthetic data reduces processing and memory resources.